There are hot startups like Granola, WisperFlow, and SpeakApp, for example, building both consumer and b2b tools to handle cases including meeting recordings, voice notes, lectures/study-related cases, manage your computer with voice, and multiple others.

But until recently, there were very few open source speech-to-text (STT) models that delivered high enough quality to use in production and build consumer-facing products. Whisper by OpenAI was the main go-to option. The first version was released in 2022 (before even the first version of ChatGPT), and the latest update was in 2023. So not much has changed in STT in recent years despite the immense financing of AI models and startups.

It’s really good news that there is a new open source STT model made by one of the big AI research companies, and the only European one. Mistral AI has released its own STT model named Voxtral. The model is available both via API and as an open source.

Mistral claims Voxtral to beat Whisper in some use cases and be on par with Elevenlabs proprietary Scribe model. Here are the benchmarks from their website.

The price for the model through API is exactly 2x lower than Whisper through API from OpenAI. 0.3 cents per minute vs 0.6 cents for whisper. So if Voxtral’s performance is indeed comparable or evene superior to Whisper, it’s a no-brainer for companies looking to build something around STT to start using Voxtral.

If you are building something in this space, please add your comments on the models, and what you prefer for what use cases. I am always happy to learn more as I myself build in this space as well.

Condensed summary

The AI revolution is in its early “inning,” outpacing previous tech shifts like the internet. Foundational AI research dates back 80+ years (neural nets in the 1940s) but only recently “crystallized” with models like GPT. We are three years into an 80-year arc. AI is ultra-democratized via cloud APIs and open-source models, sparking a Silicon Valley resurgence as talent and capital flood in. Daily breakthroughs in research and products keep surprising even experts. AI startups—from foundation models to apps—are growing revenue faster than any prior wave. Unit costs (compute, chips, power) are collapsing below Moore’s Law, driving hyper-elastic demand. Big tech, new “incumbents” (OpenAI, Anthropic, xAI), China, and startups all race to build smarter models, large and small, creating a robust ecosystem. State AI regulations risk fragmenting the market; federal leadership is needed to preempt a patchwork of harmful laws. While public opinion oscillates between fear and fascination, actual adoption (revealed preferences) is booming. Venture’s portfolio approach—backing multiple AI strategies—hedges against open trillion-dollar questions about cloud vs. open-source, big vs. small models, and incumbents vs. startups.

🚀 AI Revolution: The Biggest Tech Wave

• AI > Internet in impact—comparable to steam engine, electricity, microprocessor

• Roots in 1930s debates: adding-machine vs. brain-like computers; 1943 first neural net paper

• 80 years of AI research (“AI winters” of optimism and disappointment) until “ChatGPT moment” (late 2022)

• Just 3 years into delivering long-promised capabilities; foundational science now mainstream product

🤖 Ultra-Democratization: Cloud + Open Source

• Best AI models accessible via APIs: OpenAI, GPT-4, Gemini, Grok, Claude, etc.

• Open-source alternatives proliferate (e.g. Chinese Kȉmmy runs locally on a MacBook)

• Consumer products explode—millions adopt AI apps within months; “tokens by the drink” model

• Elastic demand: as per-unit AI compute costs fall faster than Moore’s Law, usage and revenue skyrocket

💹 Booming AI Business Models

Consumer AI

• Instantly deploy to 5–6 billion mobile-broadband users worldwide

• Pricing creativity—some charging $200–300 / mo; high prices can fund better R&D

• Startups avoid large fixed costs by tapping cloud-API models initially

Enterprise & Infrastructure

• Value-based pricing: charge % uplift from AI-driven productivity gains (e.g. marketing, service, upsell, churn)

• “God models” vs. “edge models” debate: big centralized models vs. small local models, both thriving

• AI chip boom: Nvidia leads but AMD, hyperscalers, startup designers, China, Korea, and Japan racing in

🌐 Global AI Competition

• US vs. China: two-horse race in foundational AI; Europe lags under strict regulation (EU AI Act)

• China’s open-source push (Deepseek, Kȉn, Qwen, Mòonshot); policy echo in US state bills vs. federal stance

• Federal vs. state AI regulation: need for uniform US approach to avoid patchwork of 200+ harmful state bills

• Policy engagement: A16Z advocates bipartisan federal leadership to keep US competitive

💡 Innovation Waves & Portfolio Strategy

• Venture thrives on major tech shifts; AI is the next wave after internet, mobile, cloud, crypto

• Portfolio hedging: backing multiple AI strategies—big models, small models, open-source, closed-source, consumer, enterprise, applications

• Foundational model startups (e.g. Ila, Maya, Montagna) and app-layer startups (Cursor, etc.) all vying for market

• Startups often evolve from “GPT-wrappers” to full AI model creators, backward-integrating into custom models

📊 Public Attitude vs. Reality

• Surveys express panic (jobs lost, dystopia), but revealed preferences show rapid adoption and delight

• Humans ask vs. humans do: actual behavior (AI in dating, customer service, healthcare, education) trumps rhetoric

• History repeats: new tech scares (printing press, industrialization, AI “winters”) fade as benefits become clear

🔍 Key “Trillion-Dollar” Open Questions

• Cloud API vs. on-premise/open-source: which dominates long-term? likely both co-exist in tiers

• Big vs. small models: god models for top accuracy; smaller, cheaper models for mass embedding

• Incumbents vs. startups vs. nations: who leads in foundational models, chips, applications?

• Ongoing chip supply/demand: shortages lead to gluts—massive build-outs will drive chip costs down further

🏁 Conclusions

• We’re at early stages of an 80-year AI revolution, with unprecedented revenue growth and technology democratization

• AI’s business models (usage-based, value-based) and cost collapse underpin hyper-elastic demand

• Global competition (US, China, others) and policy debates (federal vs. state regulation) will shape who wins

• Startup ecosystem—backed by diverse, portfolio-style investing—is well-positioned to navigate open “trillion-dollar” questions and thrive

• Public fear coexists with adoption; revealed preferences show AI’s real impact driving continued growth

Here’s a quick high-level summary, followed by a deep dive in bullet form and closing with the key conclusions.

• Cloud Code began as an internal CLI-based AI prototype by an engineer (Boris Cherny) and grew virally through dog-fooding and org-wide use.
• The product’s minimal terminal form factor drives fast onboarding, extensibility (hooks, custom commands), and iterative design—engineers build, ship to internal users, gather negative feedback, refine, then release publicly.
• The team is small and cross-functional: product engineers own end-to-end feature delivery, the PM sets direction, pricing/packaging, and shepherds larger efforts (IDEs, SDK).
• Feedback loops run through an opt-in 1,000-member Slack, GitHub issues with thumbs-ups, and 10 close enterprise partners—negative feedback is prized.
• Documentation lives in code (MD files, cloudMD) not Google Docs; Cloud Code itself automates issue triage, docs updates, and summarization.
• Eval regimes include end-to-end SWEBench transcripts, triggering evals (e.g. web search), and ongoing work on capability-focused tests.
• Vision for the next few months: keep the CLI the most powerful coding agent, grow the SDK/ecosystem for all kinds of agents, and make core features accessible beyond the terminal.
• Closing advice: build demos not docs, treat the agent like an eager junior engineer (give feedback), invest in cloudMD, and hone your intuition for model capabilities and realistic timelines.

🚀 Claude Code Origin Story
• Boris tinkered with API-augmenting software and built a CLI prototype.
• Cat tried it during 20% time on RL environments, loved its integration with internal tools, and began sending product feedback.
• When Anthropic decided to open-source it externally, Cat joined as full-time product lead.

🎮 Viral Internal Adoption
• Early users on Boris’s team experimented, then crossed over into research, data science, product management, design.
• High “viral factor” internally—no top-down mandate.
• Hundreds to a thousand dog-fooders provided rapid feedback cycles.

🔗 CLI-Centric Design
• Terminal interface feels familiar to any dev used to GitHub CLI, Datadog CLI, etc.
• Minimalist form factor forces ruthless prioritization of features—screen real estate constraint.
• Supports plain-text I/O only (no buttons), so every new feature must justify its footprint.

📏 Core Product Principles
• Simple first, then composable: ship a minimal version with zero onboarding UX, then expand.
• Big features get hooks, custom slash commands, sub-agents to handle different dev setups.
• CLI should handle trivial tasks immediately, but allow depth for power users.

🤝 Team Workflow & Roles
• Small, strong product engineering team owns features end-to-end.
• Engineers prototype ideas, dog-food with internal users, iterate 2–3 times, then public release.
• PM role: set high-level direction (customizability bar, aspirational feature targets), shepherd large projects (IDEs, major integrations), and own pricing/packaging for AI era.

🗣️ Feedback Loops & Community
• Internal Slack channel opt-in (~1,000 members) generates feedback every ~10 minutes.
• Public feedback via 10 key enterprise customers and occasional Twitter signals.
• Negative feedback is actively solicited; thumbs-ups on GitHub issues and direct sales-team calls highlight top priorities.

🛠️ Automation & Tooling with Cloud Code
• Slack integration: query which customers requested a feature.
• GitHub automation: duplicate/close issues, generate first-draft docs.
• Updates docs via Cloud Code’s code-aware summarization within repos.

📄 Documentation Practices
• All documentation lives as markdown in repos (cloudMD files), not in Google Docs.
• cloudMD files serve as memory: repo architecture, gotchas, test instructions, personal or team context.
• PM and designers (like Megan) even contribute small code changes and doc tweaks directly via Cloud Code.

📈 Evaluation Strategies
• End-to-end evals: run SWEBench transcripts to catch regressions, though tracing root cause often requires reading raw logs.
• Triggering evals: test binary decisions (e.g. whether to use web search) by defining clear black-and-white cases.
• Capability evals: still a work in progress (e.g. data-science tasks require gold-standard answers, large data sets, complex setups).

🎉 Anecdotes & Culture
• Easter egg: if users said “swag” or “stickers,” they got mailed real stickers. That spike flooded team bandwidth.
• To-do list feature invented bottom-up by Sid to force multi-step tasks to complete—persisted in UI after internal demand.
• Team motto: build what we wish we had, ship small, learn fast.

🕵️ Hiring & Ecosystem Growth
• Hiring a PM who loves developers, ideally ex-dev or dev-tools background, to own the SDK and grow a shareable customization hub.
• Interview asks candidates to use Cloud Code hands-on, critique it, maybe prototype a mini-feature.
• Vision of a central marketplace for hooks, slash commands, status lines—one-click install.

💡 Power-User Tips
• demos not docs: prototype features with Cloud Code before writing long specs.
• treat Cloud Code like an eager junior dev: correct its missteps on the fly, it adapts.
• invest in cloudMD: capture repo memory and your personal context for frictionless onboarding.
• use plan mode when you want explicit step lists; “plan by human demand” despite core natural-language drive.

🔮 Near-Term Vision (next 3–6 months)
• Keep serverless CLI as the most powerful coding agent, with zero-friction onboarding and deep extensibility.
• Accelerate Cloud Code SDK adoption for building all kinds of agents (legal, health, EA, finance).
• Expand accessibility beyond terminal—think Slack hooks, remote notifications, mobile usage, non-technical roles.

Key Conclusions
• Rapid, user-driven iteration (prototype → dogfood → feedback → public launch) beats long-range grand plans in fast-moving AI.
• A terminal-first tool can scale internally and externally if minimal UI constraints force clarity and extensibility.
• PMs of AI products must develop strong intuition about current model limits vs. prompt tricks vs. future model roadmaps.
• Automating docs, issues, onboarding via the product itself reduces overhead and keeps source of truth in code.
• Short planning horizons (3–6 months) align with rapidly evolving model capabilities and maintain strategic focus.

You can find the original episode here:

User Scale (MAU/WAU)

Engagement & Retention (Stickiness)

Some benchmarks on stickiness and retention from the best in class consumer AI apps: ChatGPT and Gemini

Financials & Subscriptions

And finally, some indicative numbers on revenue and the number of subscribers. Though no data on revenue for ChatGPT was presented in the original article, it’s estimated to be above $12B as per Sam Altman’s recent comments.

But if you are like me and don’t want to spend hours to get the essence of long YouTube videos you can read a detailed summary below. I made it in a couple of clicks with SpeakApp AI - best app to turn any audio content into text.

PS seems like diversity is mentioned multiple times, feels like they did not receive the memo that it’s not required anymore 🙂

Here is the summary:

Lovable, an AI “vibe coding” startup, hit $200 million ARR in under a year with just 100 employees—one of the fastest ramps ever. Head of Growth Elena Vera explains that most traditional growth tactics don’t apply: her team spends 95 percent of the time inventing new growth loops (not optimizations) to stay ahead in a rapidly shifting category. Key levers include building in public, empowering employees and users to spread word of mouth, giving away large amounts of free credits, and shipping features daily so there’s constant market noise. Product-market fit now shifts every three months as LLM capabilities and customer expectations leap forward. Lovable’s culture blends high autonomy, rapid prototyping on its own platform, generous hiring trials, and a strong emphasis on human-centered experiences—even as it pressures women’s adoption of AI.

Detailed bullet points with subtitles

🏃‍♂️ Growth at Lovable

• Under one year after launch (Nov 2023), over $200 million ARR

• Reached $100 million ARR in ~7 months, then $200 million in just 4 more

• 8 million+ users tried the platform; hundreds of thousands are paid subscribers

• Series B raised at $6 billion valuation; growth still accelerating

🔄 Evolving growth playbook

• Only 30–40 percent of Elena’s past growth frameworks apply

• Traditional optimization (funnels, micro-tweaks) gives way to big innovative bets

• Growth team focuses 95 percent on inventing new loops, 5 percent on optimization

• Need to reinvent solutions, not just optimize problems—category moves too fast

🚀 Key growth levers

• Build in public: founders and employees share daily updates, launch notes, stats

• Empower users to spread word of mouth: hackathon sponsorship, ambassador programs

• Ship marketable features constantly—small launches daily plus big “turbo boosts”

• Influencer marketing surpasses paid social in ROI; customers as micro-influencers

📈 Activation & retention

• Core product team (AI agent engineers) own activation—growth team barely touches onboarding

• Activation embedded in model improvements rather than frontend tweaks

• Paid retention on par with top B2B SaaS; net dollar retention >100 percent via upsells

• Engagement retention prioritized as leading indicator for long-term revenue

🛠 Product-led speed & feature velocity

• Growth team launches core integrations (Shopify storefronts, voice mode)

• Minimal Lovable Product replaces MVP: focus on “wow” rather than mere viability

• Full-stack prototypes and mocks built on Lovable, then handed to engineers

• Hiring “product engineers” who code, prototype, and announce their own features

📣 Marketing & building in public

• Organic marketing has shifted from SEO to social (X, LinkedIn, TikTok, Instagram)

• CEO and team share personality and challenges—no corporate scrubbing

• Maintain constant “noise” so users log in to see what’s new

• Big launches get tier-1 campaigns; everyday updates drive resurrection and re-engagement

🆓 Generosity strategy: giving away credits

• Remove entry barriers: free credits for hackathons, events, demos

• Treat LLM usage cost as a marketing expense, not margin drag

• Sponsorships for user-led events: zero questions asked on how much they need

• Free usage drives initial “wow” moments and expands paid base later

🤖 AI-driven culture & hiring

• Core question: “What can AI do here?” before adding human labor

• New role: full-time “vibe coder” prototyper; designers and PMs increasingly VIP-coding

• Hire high-agency, high-autonomy people who thrive in chaos and rapid change

• Paid, short work trials to test passion, speed, and fit before full-time offers

🌟 Community & customer advocacy

• Discord community with hundreds of thousands of members, managed by ambassadors

• Community accelerates word of mouth, re-engagement, peer learning

• Women-only hackathons (SheBuilds) to boost underrepresented adoption

• Internal team hackathons to prototype and test new ideas

🗓 Rapid product-market fit cycle

• Traditional PMF took years; now every 3 months LLM upgrades raise new ceilings

• Must place big bets in advance, then align with model releases

• Pioneer users drive capability loops; latent majority risk falling behind

• OpenAI’s market-share dip after Gemini 3 launch shows no leader is bulletproof

🏙 Work culture & boundaries

• Rapid growth: 30 to 100+ employees in 6 months; then triple again by year end

• Extreme velocity demands personal boundary-setting, not “work-life balance”

• Elena prioritizes future-self regret test: if she’ll resent a choice, she says no

• AI accelerates work; max ideation with tools like ChatGPT, Granola, WhisperFlow

👩‍💻 Women in tech & diversity

• Despite low-code ease, estimates show large gender gaps in AI adoption

• Lovable’s user base under 20 percent women according to third-party data

• SheBuilds hackathons and free credits aim to engage women in AI prototyping

• Urgent call for schools and companies to teach AI skills to avoid talent drop-off

📍 Conclusions

• Lovable’s lightning-fast growth stems from:

– Inventive, product-led growth loops over optimizations

– Constant market noise via public launches and user giveaways

– Deep integration of marketing, product, and AI engineering functions

– Generosity with free credits to unlock initial “wow” and fuel word of mouth

• AI category dynamics:

– Capabilities and customer expectations jump every few months

– PMF is a continuous recapture treadmill, not a one-and-done milestone

• To thrive in AI startups:

– Hire high-agency, adaptable talent (including fresh grads and ex-founders)

– Cultivate human-centered, delightful experiences to stand out

– Embrace chaos, set personal boundaries, and keep learning or teaching AI

• Diversity challenge: urgent need to bring more women into AI prototyping before gaps widen further.

Here is the data from that post and the reference to the source:

• Median EV/Revenue: 2.6x

The research is done by this guy: https://x.com/alexrollupguy

And here is the full post with some more details.

Usually, when they buy companies, they let go of most of the team and focus the business strictly on monetization and revenue growth. I’ve stumble upon the interview of Bending Spoons’ co-founder Luca Ferrari where he shares how they approach the m&a. The interview is available on YouTube in Italian. I’ve transcribed it, translated it into English, and summarized it in a couple of clicks using the best voice notes app (in my humble opinion) called SpeakApp AI.

Here is a quick summary and then the full interview translated into English

Bending Spoons acquires digital technology companies, uses its proprietary platform that combines tools, processes, talent and scale to unlock value faster than building products from scratch. They focus on adjusted EBITDA rather than net income to measure cash-generating ability, strip out one-off costs, and grow profitably at 20–25% per year. Their acquisition criteria emphasize digital focus, sufficient size, predictability, and clear improvement potential. Italy-based but global in reach, they hire top talent, pay market-leading salaries, and maintain a flat, meritocratic culture.

🚀 Business Model Overview
- Acquire digital technology companies with untapped potential
- Leverage proprietary platform:
• Competencies refined over time
• Proprietary tools and technologies (over 50)
• Automation and AI-based marketing optimization
• R&D investment exceeding $100 million
- Unlock value by improving product features, monetization, marketing, and organization

⏱ Why Acquire vs. Build
- Acquisitions buy time, brand, customer base and switching costs
- Viral effects and scale advantages are hard to replicate from zero
- Starting from an existing foundation accelerates market success
- Mathematical certainty: faster, more predictable path to positive results

📈 Economies of Scale and Synergies
- Bulk cloud infrastructure deals (Google, Amazon) reduce unit costs
- Shared platform functions (recruiting, payments, analytics) spread fixed costs
- Marketing automation scales across portfolio, saving tens of millions per year
- Stand-alone companies lack scale to justify large R&D or infrastructure deals

🛠 Evolution Since 2013
- Core strategy unchanged; continuous refinement of processes and tools
- Early days: manual marketing campaigns, living-room office, founders doing recruiting
- Today: 40-person data science and engineering teams, modern offices
- Emphasis on constant improvement—stagnation equals death

💰 Financial Metrics and Growth
- 2024 revenue forecast: just over $700 million; 2025 expected ~ $1.2 billion
- Net income is misleading due to depreciation on acquisitions, stock option vesting, one-off M&A costs
- Preference for adjusted EBITDA as a proxy for sustainable cash generation
- High adjusted EBITDA margins yield several hundreds of millions in profit
- Targeted profit growth per share: 20–25% compounded historically

🧮 Adjustments in EBITDA
- Exclude one-off acquisition costs: advisor fees (1–4% of deal value), severance packages
- Strip out non-cash amortization of intangible assets
- Roll back extraordinary events to gauge underlying operating performance

🔍 Acquisition Process
- Team screens ~5,000 companies annually using key parameters
- Shortlist involves internal and external experts on product, tech, monetization
- Criteria for selection:
1. Digital technology focus
2. Sufficient size to matter (>1% of platform revenue)
3. Predictability of future performance
4. Clear gap between current management and Bending Spoons’ capabilities

💲 Price and Value Creation
- Financial return driven by dollars invested vs. dollars generated, discounted over time
- Price matters but only in context of potential improvement band
- Focus on operational upside: product, marketing, tech, organization, margins

🏷 Labor Cost Myth vs. Reality
- Contrary to rumors, no cheap-labor arbitrage: Bending Spoons pays top-market salaries
- Entry-level developer: ~€64k; mid-level (~5 years): ~€200k+; 30% stock discount
- Value created through smaller, empowered teams with high talent density and autonomy
- Revenue growth and cost optimization (marketing, tech) drive margins, not headcount cuts

🔗 Portfolio Diversity and Common Thread
- Consumer-facing apps vary widely (WeTransfer, Splice, Evernote)
- Shared “engine”: common platform technologies, recruitment, marketing machinery
- Heterogeneous external brands over homogeneous operational backbone

📈 Public Investor Takeaways
- Think independently, apply rational skepticism, stay patient
- Develop and vet a few high-conviction ideas; use scenario testing and devil’s advocacy
- Stay within one’s circle of competence; deep knowledge beats broad superficiality

👥 Talent and Recruitment Strategy
- Receive ~350,000 applications per year; hire ~150 (selectivity ~1 in 2,400)
- Offer competitive pay, equity discounts, rapid career progression, high-impact roles
- Talent density and meritocratic culture: CTO at 30 years old, rapid promotions
- Low turnover (~1% per year) due to challenging work, supportive environment, variety of roles

🌍 Workforce and Geography
- ~500 direct employees (“spooners”), ~1,000 including acquired teams
- Headquarters in Milan, Italy (350 employees); other hubs: US (200), UK (100), Netherlands (50)
- Italy chosen for abundant eager talent and potential to inspire local ecosystem

💸 Funding History
- Equity raised ~$450 million: ~$250 million growth capital, ~$200 million secondary
- Debt raised ~$1.5 billion; majority of growth financed through reinvested profits and debt
- 2021 revenue: ~$110 million; 2022 revenue: ~$160 million (net loss $12 million)
- 2023 funding valued company at ~€1 billion based on compounding profit growth

👀 Investor Perspective
- Sophisticated investors saw high profit compounding, scalable platform, and rising EBITDA margins
- Economies of scale mean semi-fixed platform costs shrink as revenue grows
- Access to capital is an accelerant; cost of capital vs. return spread critical

🔄 Impact of External Funding
- Minimal operational changes due to pre-existing discipline and passive investor approach
- Governance remains founder-led, with limited interference as long as performance stays strong

🏆 Notable Acquisitions
- Splice (2018): non-monetized GoPro spin-off; huge upside in monetization and efficiency
- Evernote (acquired two years ago): large team, structured management; successful turnaround
- Worst deal: app driven by viral growth in one geography; downloads collapsed after cultural trend faded

🌐 Global Revenue Mix
- United States: ~55% of revenue
- Italy: ~2–3% of revenue

🏠 Headquarters and Future Moves
- Milan HQ to leverage local talent and inspire Italian tech ecosystem
- No immediate plans to relocate; may add offices closer to large enterprise clients

📈 Public Listing Considerations
- Pros: broader access to capital at better terms, faster acquisitive potential
- Cons: onerous reporting, public scrutiny, potential for short-seller attacks, stock price distractions
- No decision made; likely destinations: Nasdaq or NYSE if pursued

💪 Personal Drive and Resilience
- Founder has never seriously considered quitting; deep responsibility to colleagues, investors, users
- Competitive ambition to build Italy’s most valuable company and leave no regrets
- Embraces hard work, occasional discouragement but never gives up

Conclusions
Bending Spoons’ success stems from acquiring digital businesses where its proprietary platform—tools, scale, talent and processes—can unlock value rapidly. They measure performance via adjusted EBITDA to reflect true cash-generating potential, excluding one-off costs. Their disciplined acquisition process, meritocratic culture and top-tier compensation attract elite talent and fuel 20–25% annual profit growth. While public listing remains under consideration, their focus on compounding returns, economies of scale and unwavering drive positions them for continued expansion.

Here is the full interview translation in English

So, we have the blue pill and the red pill. Let’s do a bit like Morpheus; blue pill, let’s have a very mild interview, we’ll fill our mouths with marketing, I’ll flatter you for an hour, and in the end, we won’t achieve anything. The red pill, however, I’ll go in strong, be concrete, and hit hard. Come on, let’s do a mix of blue and black. Well, I’ll eat this one anyway just not to waste it.

Great. And actually, I would start off with the super concrete. Good, by the way, it’s been ages since I’ve had it. By the way, fun fact— I asked you before, are you vegan or almost vegan? Essentially, many people don’t know this, but the red dye is of animal origin, so for all the vegans listening, take note. I made a small exception today, and that’s okay. Anyway, how would you explain what you do? What is your business model in the most concrete terms possible? And if you want to add a cherry on top, how do you make money, which is ultimately what matters.

Sure, look, we find digital technology companies, not just any company out there that we think has untapped potential because the technology could be better, the product could have more useful features, more efficient monetization, more targeted marketing, and streamlined organization. We acquire them if they sell to us, and then we try to transform them to realize as much of that potential as possible, essentially bringing them as close as possible to their ability to succeed in the market. We do this with what we call our platform, which is a bit of an abstract term, but essentially means the competencies we’ve built over time and continue to refine, proprietary technologies we’ve created, our own tools that are not available on the market, and that help us do these things better than most other teams.

We have access to talent that we have cultivated over time, which allows us to employ very skilled people, and a series of other advantages that make it possible for us to do better than not everyone out there, but many, let’s say. Okay, then the spontaneous question arises: why acquire companies, given that you have all this internal proprietary technology, when you could simply develop a product?

Well, typically, in the end, we acquire time and competitive advantages. Very concretely, again, since we’re taking the red pill, if a company has built a well-known brand, it already has a customer base, for example, which gives it scale advantages, or there are switching costs. Additionally, with many customers, the brand continues to be nurtured through positive word-of-mouth, making it very difficult to launch a product that competes and gains market share from scratch.

They might have reached that point 10 years earlier, when there wasn’t any competition in that particular segment, or maybe they were the first to go there, or at that time, they brought a significant innovation that allowed them to create a competitive advantage. Now, competing with that product would be very difficult, very costly, and likely unsuccessful. Acquiring it allows us to start from a foundation of success and improve from there.

And we buy time, meaning even if we were able to, launching a product from scratch, it would typically take many years to achieve a similar position to that of that particular company because many of the effects in play, for example, word-of-mouth, take a long time to develop. You acquire customers because satisfied customers tell others about the product, and those others use it.

This is an almost exponential effect, so it takes a lot of time to build a customer base, and time is money. Yes, I’m sure you’ve talked about it on your channel—at the end of the day, for us, acquiring means having more certainty of achieving a positive result and reaching it sooner. So, it’s a mathematical issue, and while it may seem trivial, I don’t think it is.

If these companies have this potential, why sell? Well, because potential doesn’t exist in a vacuum; it exists in connection with other factors, like who works there. For example, a company might not be able to unlock its potential because that potential can only be realized within the broader context of Bending Spoons, where we have economies of scale that the individual company doesn’t. For example, when we negotiate with our suppliers, Google, for instance, or Amazon for cloud technologies, simply the expenses of cloud infrastructure are significant costs in a digital business, and we negotiate contracts that may be 10 times larger than those of the single company we’re acquiring, allowing us to obtain much better terms.

It’s not that the people at that company aren’t good; they just don’t have that advantage. Yes, yes, utilizing scale. This is just one example of many. Another example is that we’ve built many technologies, our own tools—more than 50 over the years—that allow us to do things very well. One of these, for instance, is a series of automation tools, some based on artificial intelligence, that allow us to manage marketing expenses, especially on search channels, like when you go on Google and search for video editing apps; we can do it completely automatically. We invest tens of millions each year, and it’s an algorithm that does this, and to have that series of tools, we’ve invested significantly over the years, more than 100 million in total in research and development across all these technologies.

Clearly, this spending makes sense to the extent that we can then reap benefits across the many businesses we own and manage, but if you have only one, that’s not an investment that yields a positive return. This is another example. There are many others. The fact is that very often, the company on its own is not able to unlock this value, and sells to us because we can unlock it.

It’s a win-win. Yes, yes, yes. And has the business model always been this way? Has it changed over time? I mean, how did it start? What were the origins compared to what it is today?

I’d say in the first person it’s always been the same. So, as an individual, not the company I represent today, I ran a startup before Bending Spoons from 2010 to 2013 that had nothing to do with this strategy; it was a classic product idea, in that case, it was a diary that had to be written by itself with AI and that went poorly. But since we started Bending Spoons in late 2013, the strategy has been this.

It’s clear that more than 10 years have passed; stubborn people do not change or improve. We’ve refined many things, enhanced, improved, but the core has remained the same. And regarding that, for example, you said you’ve refined, improved, are there any specific things where you feel like, “Wow, we’ve really changed a lot”?

Well, then the metaphor, since in the end, a nice metaphor is sports, right? I mean, if you play tennis, you start at 12 years old training, always playing tennis, and there are the forehand, the backhand, right? When someone becomes Novak Djokovic, they find themselves playing Wimbledon; the strokes are always the same, but it’s all 10 times better. Our journey has been somewhat similar conceptually; almost everything we do is better, from frankly not very significant things, like the office.

Today we’re recording here from our office; when we started, we were in the living room of a shared apartment with the other founding partners. I mentioned earlier about that technology, those marketing automations. At the beginning, it was one of my partners who literally manually handled all the campaigns, checking the data, and the recruiting team. Initially, I was the one downloading lists from Alma Laura and calling one by one all the graduates, let’s say, from the Polytechnic of Milan, the University of Padua, trying to convince someone to come work with us.

Today we have a team of 40 people with data scientists and research engineers creating predictive models, lead generation campaigns, so really, almost every area has improved. Luckily so much because we work incredibly hard; it would be dramatic if we hadn’t improved. Well, actually, for the Italian market, you often have companies that stay exactly the same for 10 years, and that’s something that, in my opinion, needs to be understood—that if you stop, you die.

This is a concept that I believe needs to be conveyed. But let’s get back to the concrete; let’s share some numbers. I think 2023 isn’t closed yet, so let’s talk about 2023. A few numbers, such as revenue, profit. 2024. Okay. Fine, 2024. Look, in 2024, I’ll give you the numbers in dollars because internally we deal in dollars. Still, the euro is close to the dollar; it changes a lot. We expect revenues of just over 700 million, and for 2025, we expect revenues probably around 1 billion and two.

In fact, we don’t look much at the net income— as they say, profit margins— because valuing a company from the outside that may not have other, it’s fine, but for a company like us that is very acquisition-driven, net income tends to be very misleading as a measure for a series of reasons. Then I’ll tell you a measure of profit that we use instead, which I think is more adequate, at least in our case. Net income tends to be misleading because there are many reasons, but let’s say, assuming we’re looking for an indicator that gives us a sense of the business’s ability to generate cash looking forward, typically one seeks that reasoning.

First of all, when we make an acquisition of an asset, there’s depreciation, and that depreciation there applies, depending, but maybe for 5 years, it kills net income. But if you decide tomorrow to stop making acquisitions, that’s not really a cash impact—an investment you made years before—of an asset that might last 20 years, 30 years generating you cash. However, regulations tell you that you must depreciate it over a fixed number of years, and this kills the margins on net income.

Another thing you can have are all sorts of aberrations or accounting artifacts due to the fact that you have to consolidate a balance sheet. We have many subsidiaries, many companies we’ve acquired, and we will have 30 that we try sometimes to liquidate, but we’re always acquiring new ones. You must, like mother Bending Spoons S.p.A., consolidate all these balance sheets, and you end up presenting things that are truly distorting.

For example, when we acquired one of the companies in 2024, as is very typical, both the management team and all the staff had stock options, and as per how the stock option plan was structured, in case of acquisition, all these stock options that had a vesting period in the future would vest immediately, of course. So, we paid a price that had nothing to do with this aspect that was what we were willing to pay, and the seller then knew that they had to redirect some money, in this case about 90 million dollars, to all these management team staff for these stock options.

However, when you aggregate, let’s say, consolidate the balance sheet, you find a cost line of 90 million in 2024, which is completely an artifact in the sense that it is neither indicative of the cash generation of that business nor of your business looking forward. This is tied to the transaction event, and it also hides the fact that, in reality, that cost was actually borne by the sellers, who accepted a price that was essentially a bit lower to then compensate employees and management according to the agreement they had with them.

So, that thing kills the net with that 90 million, but it has no relevance on your ability to generate cash. Or extraordinary M&A costs. When you do an acquisition and you have an advisor, it is perfectly normal to pay 1, 2, 3, 4% on acquisitions of 500 million, and you understand that can also translate into tens of millions. So, net income tends to be a bit misleading.

We look very closely at EBITDA or adjusted EBITDA, which is essentially the operating margin, that’s a better proxy to determine how much cash this business generates if we stopped making acquisitions. That value we don’t share externally, but I can tell you that our margin is very high and, in 2024, several hundreds of millions. On the other hand, adjusted EBITDA, or adjusted, okay?

And what’s the difference in the adjusted one? What do you take into account? What don’t you take into account? The adjustments, the main adjustments in terms of impact are related to one-off acquisition costs, for example, the costs of advisors. What we often do is typical for us; when we make an acquisition, we do a mass layoff if we believe that company will function better with a smaller team. We provide generous exit packages well above the market, and the cost can be substantial. Last year, several tens of millions.

That again is something that you only have one-off after the acquisition event and is not indicative of a recurring cost that the business will have. So typically, when you go to make adjustments, this is something that auditors also do, I don’t know, from EY or KPMG, and you would consider it an extraordinary cost that you roll back to arrive at these adjusted EBITDA, which still amount to hundreds of millions of dollars, but with significantly higher adjustments. This is also the reason why the last round that just concluded valued us at almost 5 billion euros, and these are certainly very sophisticated investors, not someone who says, “Ah, they don’t know anything.”

They wouldn’t do it if they didn’t see that ability to generate cash. Now I stay on this topic because I’m very interested—also because I think you know—one of my models is Warren Buffett, who has his own opinions on EBITDA, but there we’re talking about public companies, it’s a different dynamic.

Well, I agree with him, by the way; I think it depends. Okay, okay, anyway, I’ll let you finish. Well, the dynamics are a bit different. Let’s say I think a private company should be evaluated differently—that it shouldn’t be evaluated the same way. However, for example, you mentioned revisiting things like advisory costs or severance packages, but if that’s part of your business model, can’t it be considered instead as something that I expect to happen regularly, so I do, in fact, consider it a cost?

Yes, it depends on what you’re looking to evaluate. First of all, let’s take a step back. Warren Buffett says that if you have a business that continues to require capital investments—well, he created his theories in an era when investments in industrial goods companies were predominant. If you have a business that requires ongoing investments in what is called property, plant, and equipment, then you need to be careful.

If you focus only on the operating margin, you miss the fact that that business will continue to consume cash, or if you’re a business that has a lot of inventory—so as it scales, it continually sucks more capital to increase working capital. You’re missing a significant negative profitability factor, and in that case, EBITDA would be misleading.

For us, these things are zero. Essentially, we don’t have inventory, we don’t have significant investments in any asset that involves depreciation, amortization, except for an intangible asset, like IP, exactly. So it’s quite a different thing. We also have a very unusual model, whereby there are very few businesses like Bending Spoons on the market. If you’re trying to assess profitability as it stands today—if you’re trying to evaluate how much cash our business generates today—then I think excluding those one-off costs is right because then you say, “Okay, let’s pretend Bending Spoons stops making acquisitions as of tomorrow; over the next 10 years, how much cash does that adjusted EBITDA generate?”

That gives you that answer. If, on the other hand, your goal is to understand what the cash generation is in that specific year, then absolutely, adjusted EBITDA is not the right tool, nor is net income, because there are a lot of non-cash costs, like I said before, the amortization of an intangible asset. In that case, I think a more interesting question to ask is how quickly I think this business could produce the famed compounding, so you choose a profitability measure that you think works for you, whether it’s adjusted EBITDA or net income, and you ask yourself how quickly you think this could grow over time.

We historically maintained—this was also an indicator Warren Buffett cited in his letters to shareholders at Berkshire Hathaway—historically, not so much in recent years, too large—but it has always been able to average between 20-25% growth per share. That’s a very interesting indicator, and you can evaluate it, if you’re not publicly traded, you can evaluate it on net income, adjusted EBITDA, just EBITDA; whatever one you look at, averaged over time, we’ve grown extremely fast to record levels, much higher than 20-25%, being even smaller makes it easier.

I’m saying that we’ll reach 20-25% valuation at a trillion like—well, so if you look at it, if you want to assess Bending Spoons on its current status—saying this perimeter today, if they stop acquiring, how much they earn—then I think adjusted EBITDA is the closest thing to optimal for doing that, and you get an answer. If you want to disregard how much they decide to reinvest back into the business—because they make acquisitions, pay advisors, etc.—how quickly does this indicator of profit increase over time is another way to ascertain how much value there can be.

So whether you choose EBITDA or net income, ultimately you reach a similar answer, even though in a particular year, one can be favored while the other is disfavored, but if you look over a period of maybe 2-3 years, the answer you reach is quite swiftly. Excellent.

Let’s talk a bit, go back, and discuss the process that leads you to say, “Okay, I see potential in this product, in this company, I evaluate it for acquisition.” I’d like to know what you originally see that makes you say interesting and then how the process actually works briefly. Well, there’s a team of about ten people trying to evaluate as many things as possible. Last year we considered over 5,000 companies, even superficially, and they have a whole series of parameters. Now I can’t say everything, but several parameters through which, let’s say, we narrow down this list so that it’s more manageable; the more we approach a shortlist, the more we involve experts from outside this team who work at Bending Spoons and who offer specific contributions on how something can be improved, how monetization can be improved, how technology can be improved.

We refine, and we have projections essentially, and at the end of the day, the criteria for selecting a company are digital technologies. Currently, we don’t evaluate supermarket chains, even though they could be interesting to consider in the distant future. Well, since Amazon wants to make the cashier-less ones? Exactly.

The size must be large enough that it shifts the balances, because as I said, this year we expect revenues of over 1 billion. It’s clear that if something adds a million in revenue, perhaps in relative terms it could be very profitable, but it results in more headaches than it’s worth. Third, we must be able to accurately predict the future for that business.

So there are some areas where we don’t feel comfortable. For instance, we’ve never made acquisitions in gaming where, perhaps if you consult an expert, they’d say, “No, Luca is wrong,” but my point is it’s not that I understand gaming; in fact, we understand little. It seems to me that the trajectories of these businesses are a bit based on current trends, and so we tend to be uninterested. But we can take a look at everything; however, it’s essential for us to make an investment knowing what we can achieve.

The final criterion is we must be convinced we can manage that business much better than how it was managed before, and that’s the only way for us to make an attractive offer to the potential seller while thinking we can generate interesting returns for ourselves. The broader the range of trajectories of a business managed by those currently handling it versus how we manage it, the wider the band, the more this is true, and thus we can make everyone happy, essentially.

And when is that range wider? Again, we must think while analyzing that business from the outside that the product can be improved, the organization can be improved, the monetization is far from optimal, the marketing is far from optimal. We try not to share the exact indicators because, naturally, that’s a bit of a competitive advantage.

Of course. How important is the price paid for you? Well, I mean, ultimately, the purely financial side boils down to whether I invested a dollar or how many dollars I generated over time, you know? Then with the appropriate discounts for time in the future, what you want, and clearly in that equation, price is a factor.

Okay, apparently the camera overheated. I hope that someday you will buy Sony to fix this issue too. In any case, the meaning is clear: there must be potential for improvement. Perfect. Before preparing this interview, I read not only past interviews and listened to those done, but I also looked at the comments under these interviews, and I found one that recurs—not too much, but I found it interesting because it summarizes, in my opinion, both a mentality and a legitimate doubt that might arise, namely: what do you say to those who claim that all you do is buy a company that may not be doing well and fire a lot, hire cheaper people, and there’s the profit?

Uh, there’s a lot to say here. Let’s see where we start. Well, first of all, it’s not true that we necessarily take companies that aren’t doing well. We’ve taken high-growth companies, companies that were stagnating more or less, companies on the decline—everything, so that’s not true.

Secondly, it’s also not true that we profit solely from the per capita cost of the staff. That’s wrong. I’ll give you the data, the facts. For the same role and seniority, we pay at the top of the market in Europe, and it matters that we take the role of software engineer, which is the most represented at Bending Spoons but also at the companies we acquire by far. A recent graduate with us typically earns €64,000; a mid-level software engineer, with around 5 years of experience, is in the range of—let’s say around €200,000 or more.

It’s worth noting that this is all fixed. When you look at the headline numbers in job offers, there might also be parts with vesting over 4 years, variables that are unclear, so everything is fixed; that has a value, and one has the opportunity to invest in Bending Spoons stock at a 30% discount, so when you factor that value in, the numbers I just mentioned should be further increased. So, not only are these values completely off-market in Italy, but they’re also at very high levels in Europe.

We often hire people from top universities, including Oxford, Imperial, Google, Meta, so clearly we cannot pay at mediocre, let alone low levels. When we’ve made acquisitions where there was a workforce in Europe, the per-capita cost in the acquired company was always lower than that of Bending Spoons; in reality, the phenomenon was opposite where the workforce was in the United States, which generally cost a bit more, but not double; that’s not where we made our profits.

Where do we generate value? Let me discuss costs and revenues. Starting with costs, where we generate value is typically definitely not paying people less; on the contrary, we often pay more, by creating smaller teams with fewer managerial layers, much more autonomy, greater talent density, unleashing essentially the sense of ownership and entrepreneurship of people, doing more with smaller teams, I reiterate, often better paid than before, and then optimizing marketing expenses. Sometimes we find large budgets that are maybe managed inefficiently in channels or initiatives that don’t bring value. We’re, I believe, good at being very rational and scientific, I would dare say, in optimizing that.

In terms of revenues, we generate value mainly by improving the product, enhancing new marketing strategies, acquiring more users, improving monetization, and enhancing technology. So that idea of arbitrage on labor cost is completely false. It is absolutely true that we often cut a significant slice of the acquired team, but this is not because the business, as you see, was not doing well; in fact, many of the business acquisitions were growing, but we simply believed that a smaller empowered team, with more responsibilities, less process, less management would do much more than a bigger team with more bureaucracy and processes, and I think the facts have proved us right.

Perfect. Going back to your products, I noticed that they are very different. In fact, I don’t see a common thread because I’ve seen apps like WeTransfer, and I’ve seen products like Splice, which is for video editing. Is there actually a common thread, or is there perhaps a willingness to create this heterogeneity to diversify? We like diversification, but it’s not something we actively pursue; simply, the common thread is, look, to give you another metaphor, it’s like they’re all cars with different liveries, different brands, but with a common engine; what is common is the way we work, the technologies that underlie them, that platform in the strict sense I mentioned earlier.

Now, what we call the external image that the consumer or the company buying our product sees tends to be very heterogeneous; we don’t pursue consistency from that perspective. I think an interesting example in a completely different market is P&G—Procter & Gamble, which I think your audience may know. They have many of the most famous supermarket brands in the world; often, consumers don’t even know they’re behind the same company, and what’s common is that operational and marketing machinery, but what we have in common is technology; hence, the market is completely different. Okay.

So, it’s not so much about the nature of the problem that the product solves but more about the organizational base. Okay? We’ve somewhat explored your acquisition method; your process is very function-based on what you’ll do after. Yes, but I’m speaking to an audience that mostly focuses on publicly traded companies, so it’s a whole other league, and they can’t directly influence changes to the businesses. So, just to say, is there something you’ve learned from this experience that you think could apply to someone wanting to invest in Procter & Gamble, for example?

Without a doubt, but my advice is always to, when making investments, think of yourself as being less skilled than you would naturally believe, so buy an index—then, I don’t know—because there’s a world of well-armed professionals in investments; it’s hard to outdo them. However, at a high level, then maybe we can double-click on a couple of things. I’d say that rationality and independent thinking go hand in hand, and then patience. By rationality and independent thinking, I mean being able to listen well to everyone’s opinions—everything you find online is correct, important, indeed stimulating to think. But it’s essential to form your own opinion, in my opinion, with your own mind, to be very critical, skeptical, and to question things, going deep into them; this saves you a lot of mistakes.

Typically, the best investments are also those with a somewhat counter-cyclical, somewhat controversial edge, and if you listen to mass opinions, it’s hard to identify them. Well, in terms of patience, I’d say that in general, emotions, impulsiveness, enthusiasm are poor advisors when it comes to investments, in my opinion. So if an idea comes, never act on it immediately—you know, no one thinks they are going to be a high-frequency trader who, if they don’t catch a stock within 5 seconds, should never act on it immediately, in my opinion. Note down the idea, look at it again after a few days, a week, if you still like it as much, that’s a good sign. If the emotion has passed, and you say “Ah, that was a bit of nonsense; maybe I won’t do it.”

I’d also add that it’s useful—in the sense of conscientiousness and discipline—to invest time in refining a few ideas rather than trying to have many or pursuing many. For example, I suggest writing down an investment idea, describing why it’s a good idea, why it’s not a good idea, showing it to someone else or if you don’t know anyone with a minimum of expertise, to try to wear the critic’s hat yourself and dismantle it in every way, I mean really taking the time to create a process almost like a natural selection of ideas, where if an idea survives on the other side of that tunnel, it’s probably good, it’s anyway much better than the average of those that were at the beginning of that process.

And finally, I believe that competence is much better in life, certainly in investments; knowing a lot about a few things is preferable to knowing a little about a lot of things, provided that a minimum general understanding always helps, so going deep, studying a topic well, doing your homework, and accepting that you can’t do everything. Well, Warren Buffett, who is also a personal hero of mine, and you mentioned him, always said that it’s important to stay in, well, your circle of competence, right? So these things we see very much even on our end. I think they’re true for any investor.

All of them seem to be excellent tips. Well, since you were born, you’ve received hundreds of thousands of applications, and one of my questions, and I ask you this as an entrepreneur who also deals with people—what’s the secret, if we want to say, even if there isn’t one, to being so appealing in the job market? Is it just paying people a lot?

No, well, certainly offering very competitive pay is an essential part of this for nearly everyone. It’s a top-three factor, let’s say, at least. It’s probably not about maximizing remuneration but feeling that one is compensated at least in line with market expectations. It depends on who you want to attract, in my opinion. I see jobs as a product in the end, and thus it must be improved according to the customer you want to buy it, in a sense.

So, I can tell you the kind of person we look for and for whom we’ve tried to optimize work here; it’s a person who is very, very ambitious, who genuinely seeks to learn a lot, who wants to feel that they’re quickly taking on responsibilities and having a substantial impact, and advancing rapidly in their careers; they are clearly very talented individuals. And for those people, what we aim to offer, which I think is a bit of a “killer feature,” is, in no particular order—aside from very competitive compensation—an extremely high-quality colleague base; we talk about talent density. Last year, more than 350,000 candidates applied to work here, and we hired 150, so it’s one in around 2000-2400, which is a selective rate; it’s the most selective process I know, certainly more than the famous Big Tech companies, more than Stanford, Oxford.

And what happens is that when you have that level of selectivity, you work with very strong individuals, which is fantastic because you learn a lot, and the work itself becomes very satisfying. This is an extremely important factor. We have a very flat and meritocratic culture that allows a talented individual to advance at a speed that is totally off the scale compared to most companies. Once again, since we took the red pill, let’s give very concrete examples: currently, our CTO, the person who leads the entire engineering and technical side—who also leads all the acquisitions, which means well over 500 engineers—is just 30 years old and has been with us since just before college graduation.

A role like that, if you look at the market, you find people of 50 years old with 20-30 years of experience. I believe he’s very competent, even more competent than those who’ve earned it in the field. Since we’ve had the so-called courage to give him this responsibility very early, even the person who leads Evernote, there was a CEO previously, around 50-60 years old, and now she’s about 28. You don’t find many companies like Bending Spoons that if you’re very deserving, they leap and give you all this responsibility so early?

So, this is the second factor. The third factor is a lot of freedom, flexibility, trust, and a very collaborative atmosphere, making the environment, I think, enjoyable to be in. Obviously, there’s pressure because we want to be super high-performance, but it’s also a very supportive, friendly, informal environment, and consider that we have a turnover rate of about 1% a year, which is very low in the tech world, rarely less than 7-8%, so an absolute outlier, and the things I mentioned earlier are an essential element of this too.

And the last point, which is something I realized only in the last year or two, so it wasn’t by design as they say, but it has become important, it’s very cool; it’s been a realization for me. It’s really cool to work in a place where without changing companies, so the rules of the game, colleagues are more or less the same, you can face very different challenges in very different areas. For instance, you might spend a year on Evernote, then six months with WeTransfer, then work at a platform level on a payment management tool, for example, and that’s not something that many companies, almost none, offer.

So, I think this is another element in terms of talent for us of competitive advantage, without a doubt. Today, what are some workforce numbers? How many are there? How many are in Italy? Are they all in Italy? No, no, we’re not very distributed. So, at our company, which we call—let’s say, spooners, people we’ve hired directly—there are just under 500 of us.

Then if we look at all the acquired teams as well, we’re around 1000, roughly speaking. Concerning geographic presence, anyway, Italy is the primary location; I’d say we have more than 300 people, about 350 in Italy, around 350 of the 500 in total. Practically all of the 500 are here. Then we probably have about 200 in the United States, around 100 in the United Kingdom, 50 in the Netherlands, and then a long tail of various countries, mainly in Europe and North America, but clearly.

You’ve received $340 million in funding in 2022, plus new rounds, and in 2022, your revenue was $100 million. So my question is, what did those people, those institutions that invested in you see to justify a multiple, a potential investment so high?

Yes. Okay, let me clarify. In 2021, we made $100 million, okay? $110 million, I think, with a profit of $15 million, net profit, the bottom line. In 2022, we made 160 million and lost 12. Okay. Still, the concept is valid; I just wanted to clarify that.

Perhaps I should also clarify the funding we obtained so that we have all the—I think you like numbers as much as I do. So in the first capital increase, we raised around $450 million in equity, of which about $250 million was in capital increases—so injections that finance the business—and about $200 million in secondary funding, meaning investor X selling to investor Y to liquidate their shares.

The company is completely neutral in that respect. It’s simply people who want to cash out and people who want to invest. Concerning debt, we’ve raised about $1.5 billion. Historically, almost all funding has come from reinvested profits and debt, with equity around $250 million, which isn’t much; it’s not little, of course, but it also isn’t preponderant relative to the figures I’ve mentioned.

In 2022, most of that figure you mentioned was debt, there was about $50-60 million in equity, almost all secondary, so we did not finance with an exit. The first significant equity capital increase investment was in 2023 when Bailey Gifford, Cox Enterprises, soon after Durable Capital, came on board with about $70 million in capital increase, about $50 million in secondary funding, and this was also the first time we were valued at around a billion euros.

What did they see? Well, those investors clearly had access well beyond the public balance sheet; they were able to make all the necessary adjustments and due diligence, they could see our projections for the future, analyze thoroughly. Essentially, they saw some of what I mentioned earlier, so a lot of that net income is muddied by the fact that there are many costs or accounting aberrations or one-off acquisition-related costs, so it doesn’t give a full representation.

In fact, those years, EBITDA was around $20-30 million, which by the way was already more in line with that evaluation between half a billion and a billion. In reality, in 2022, as you said, we were worth less than half a billion, with BIG EBITDA of $20-30 million, which is perfectly in line with the multiples on the market, but above all what they saw was the speed at which we compound, that is, how much profit increases each year per share, which is what they care about.

Then there’s all the dilution effect, of course. Exactly. Additionally, we had a track record that improved in relative terms even in terms of diluted earnings per share; we grew faster in 2023-2024 and still expect to grow in 2025 at a rate exceeding what we did over the last 23 years, so they saw that aspect, and they said, “If these people continue to perform around the past numbers, I can think about where my capital will go, assuming that the multiple remains relatively similar; I can imagine my capital will grow at that rate,” and it was a very attractive rate for them.

Furthermore, keep in mind that our model benefits from economies of scale, in the sense that there’s a semi-fixed cost of operating this platform. All these teams are not specific to any product but work on these technologies, tools, services that all products use, from recruiting to processing all the payments, to the marketing automation I mentioned before—there are over 50 in total, as I mentioned.

Everything here is relatively stable over time. So, the more revenue you generate by adding businesses, the smaller the impact of that cost on margins. For instance, if back then the EBITDA margin was about 15%, so okay, nice, but nothing extraordinary, nowadays it’s much higher, and we expect it to continue growing in the future precisely for this reason.

Obviously, these investors are sophisticated; they were able to look under the hood and convinced themselves that this would happen. When you receive so much funding, whether equity or debt, in general, someone is betting on you; does it change how you operate? Are there added pressures that lead you to behave a certain way? How does it work?

A little, a bit of it changes. How much it changes depends on a variety of factors: one, how structured, rigorous, and disciplined you were beforehand. The more you were, the less you need to change. The more amateurish you were, relying on gut feelings with little reporting, the more you need to change because investors expect you to inform them about how things are going.

We felt we were already quite structured, professional, precise, and disciplined, so the increase in complexity was relatively contained. It also depends on the type of investor or creditor that you bring on board. We’ve had very passive investors—I say this positively—meaning they observe, but their mandate is not to pressure you, unlike some others, which is also a perfectly sensible way to operate, depending on the type of investments you make.

Moreover, we’ve done well, and you can imagine how the cracks appear, particularly when things aren’t going well. As long as you’re doing well, they say, “Well, a winning team doesn’t change.” I must say we have always managed to maintain a solid governance structure at the statuary level, so the margin for interference from these partners is still very modest. However, one can exert pressure, even just psychologically, you know?

So, in relation to the acquisitions you have made so far, is there one in particular that you think has been the best either in terms of metrics, numbers, or perhaps even for what it has taught you?

Yes, I tend to think that the next one is always going to be the best.
I imagined that the red pill categorically forbids that.
Yes, look, I’ll mention two that I think have been quite watershed moments for us. The first perhaps is Splice, acquired in 2018, a video editing app that was part of GoPro—speaking of cameras, overheating. Exactly. It went very well because it wasn’t even monetized; being a very small business for GoPro, they hadn’t taken care of it much—understandably so.

At the time, we were much smaller, so for us, it was significant. Not being monetized, we succeeded in making extraordinary economic improvements, which is typically unrealistic, but it was. It was also the first time we acquired from a very structured institutional seller; until that year, I repeat, in 2018, a lifetime ago, we only acquired from, I don’t know, individual developers from Canada who had a small app, which involved a completely different shade of negotiation—bureaucratic and legal dynamics.

So, nonetheless, it was a crucial shift for us. The second one I’ll mention is Evernote, acquired around two years ago, which was the first time we acquired a company that already had a significant team, a very professional management structure with very credible track records, so it was a pretty robust beast. We had never done that before, and I believe we were somewhat understandably insecure about whether we would be able to do significantly better.

The answer was yes, and it wasn’t obvious—this gave us great self-confidence and also a lot more credibility regarding various financial investors. In fact, the most significant rounds we conducted happened about 6 months after the acquisition when it started to become clear that we were doing good work. So it was also a sign of saying, “Hey, we are serious about this.”

Yes, we’ve always had that; consider that when we started 12 years ago, we said we wanted to build the best company in the world, which then, of course, is entirely illegal—the red pill entails saying such things—but I mention it only to convey the level of ambition we had. However, it’s clear that someone on the outside, when they see you making a big step, believes more if they see young kids—maybe back then, I don’t know, 28-30-year-olds saying, “We can pull this off.”

Sony, please grant us the last 10 minutes. In any case, clear. Therefore, Splice and Evernote for the reasons stated earlier. However, now it’s also your turn for the other part; what’s been the worst acquisition? And no, you can’t just say no, they were all fantastic.
No, no, because by definition, “worst” means the least good. Absolutely, I wouldn’t dream of it. Look, let’s say this: we make many errors, many blunders, we’re very imperfect. Among these imperfections are acquisitions that have gone worse than we hoped. There’s one that immediately comes to mind; I won’t mention the name out of respect for those involved, but it was an app that when we acquired it—speaking of acquired companies only at difficult times—was on an exponential growth trajectory, and immediately after we acquired it, but even before we had time to intervene, the number of users downloading it plummeted.

Yes, in that case, it was download numbers, how many sought and downloaded the app. Before this reflects on revenues, of course, there’s a time lag because all the active users and subscribers need time to feel it, but it’s an important indicator—this was several years ago—what we learned from it is that that exponential growth was primarily driven by a viral phenomenon in a particular geography. Specifically, if I remember correctly, it was in Indonesia, where it had essentially spread socially, so many people had started using this app tied to this social phenomenon.

And once that viral social phenomenon waned, naturally, the number of people looking for the app sharply declined accordingly, which significantly reduced the value we assigned to that product, and it turned out to be a rather mediocre acquisition that we would not have made had we known this fact. However, wanting to look at the silver lining, it did teach us a lot, and now we’re extremely meticulous in investigating all the traffic drivers of a product we buy; we want to understand who is searching for it, why, from where, to isolate any phenomena such as this and ensure the quality of the traffic we’re purchasing.
Yes, so it must come from healthy growth.

Yes, exactly. Or else we evaluate it less—meaning we could also consider it; we wouldn’t have paid as much as we did grand unless it appeared justified. Certainly! And is this still an app you have?

Yes, we still have it. It does just fine, but looking back, I absolutely wouldn’t do that same acquisition. Okay, that’s clear. Currently, the bulk of the revenue, if I’m not mistaken, comes from outside of Italy, right? I believe the United States represents around 55%.

Okay, Italy will be 2 or 3%. Perfect. Have you ever considered delocalizing, that is, actually moving abroad, or are there specific reasons for staying here?

Well, we still have people, as we discussed earlier, from various places worldwide, but the headquarters is in Italy, in Milan, and as I said, the predominant share of people is in Italy. So, there isn’t a particular reason for us to move, at least not based on where our clients are, because most—well, it’s all digital, and there’s no need to be situated in the same place.

Certainly, as we continue acquiring more products that serve large companies—like we acquired Brightov, which was publicly traded before and is now private because we bought it, which has 2,000 clients, all big companies—then it might be convenient to be in the same time zone, simply so you don’t have to wake up at night, or maybe you want to meet in person. So there will be situations where we might want to grow the organization somewhat where clients are located. However, historically, this has not been that important.

We’re in Italy, primarily in Europe, for two reasons. It was a very conscious choice, and many people don’t realize that we started in Copenhagen, Denmark. We moved to Italy after about a year and a half for two reasons. First, we evaluated that there were many brilliant and eager young people in Italy, but relatively few real job opportunities where they could be very ambitious and hope to build something really cool.

Thus, it was a win-win situation: we could build a stronger team, and those people could avoid the hassle of moving abroad if they didn’t want to do so while still having a stimulating career. The second reason was that we said if we were to fail, the place of failure wouldn’t change much to us, but if we managed to create this fantastic company of world caliber, doing it from a country like Italy, which has seen very few successes in the last 20-30 years, particularly in technology, would have a substantial value ripple effect in inspiring other entrepreneurs, maybe even companies that exist already, to be more ambitious, to make a leap in quality, perhaps initiating a small or even large economic and cultural revolution.

You look at Silicon Valley; it wasn’t always like that; at a certain point, in the early 1900s, it was a very agricultural place, not particularly futuristic, and then a series of coincidences led to the founding of Fairchild Semiconductors, which became the mother of the semiconductor industry, and from there, a chain of capital, know-how, and new businesses emerged, and that cycle repeats itself. Today, it is the primary hub for innovation and economics.

So, I’m not saying Italy will become Silicon Valley, but if we can be a great example, we might make a few more steps than we otherwise would. All these factors remain true today, just as they were back then, so we don’t plan to move our headquarters. Perfect.

The subject of going public is an issue. What? Look, we’re always asked about this; it’s discussed constantly, and I imagine there are some people who occasionally knock on the door saying, “Hey, I’d like to make an exit.” That has never been a problem because we organized, as I mentioned earlier, that we’ve raised $450 million in equity, of which almost half was $200 million secondary, meaning that shares changed hands.

We’ve always managed to find buyers for anyone wishing to sell; therefore, we’ve been fortunate enough to do well—and there’s always been strong investor interest. So, when someone wants to cash out, we help them find buyers. In the end, it is slightly more cumbersome than if we were publicly traded, but it’s not a significant problem. Sure.

The advantage of going public for us would be access to more capital with generally better terms; well, let’s just put aside that, it can indeed make a difference, especially in a highly acquisitive business like ours, where if we have access to more capital, we can invest at very high return rates as long as the cost of that capital is significantly lower than the returns we generate—it’s a significant accelerant.

The disadvantages are—it incurs even more onerous reporting; you have to disclose things that make it easier for competitors, of course. There could be significant media pressures. There’s a strong incentive to speak about publicly traded companies, and while it could be malicious, there’s an entire world of short sellers, you know, which one person would have you know, lately, going very badly.

But yeah, if Waters holds, they’re likely closing, so there’s a percentage of these who have no scruples and spread untruths about the companies to try to lower the price, and if you’re the company, you find the untruths turn into reality when people worry and call you. You have to manage that; it saps energy and time.

Then there’s the aspect of having a stock price that fluctuates over time, which can be highly distracting. It’s not easy to say, “Well, just forget about it, we watch it once a year.” Yes, it’s easy to say to others, isn’t it? That’s the problem. Exactly. So all these negative effects make it unclear what will prevail. As of today, we are very uncertain about whether to go public or not. We could; we have many proposals from investment banks to do it, and we are always evaluating them, but there is no concrete plan. If we were to list, the most natural choice would be Nasdaq. I would say that, and perhaps in second place, I might say the NYSE, so the New York one. However, if we really had to work seriously on an IPO, we would need to evaluate all options, and it’s not a taken-for-granted decision. Okay, but I mean, I would have been surprised if you had said, “Ah, you should list on the Italian stock exchange.” There are issues, and your response further confirms to me that it’s challenging to have a company listed in Italy, especially when it comes to attracting capital. Yes, from a technical point of view, it would be easy, perhaps even easier as an Italian company, but there are some extra complexities for us to list in the United States. Well, simply put, the multiples are generally a bit worse because the demand-supply ratio is slightly more unfavorable when you list there. It’s hard to ignore this aspect, right? Yes, definitely too important. Next acquisition? Just kidding. Um, but I don’t know. No, no. Uh, we have one that should close next week. Alright, we can talk about it off-camera, right? I want to wrap this up. So, we’ve been super concrete and I really appreciate it because it was exactly what I hoped for; you’ve been great. Um, but let’s close with a moment of inspiration, I think we also need this little moment of blue sky thinking. Was there ever a moment in your career where you thought, “Maybe I can’t do this, maybe I should stop, maybe I should give up?” And what kept you going? Obviously, because you are here. Well, I mean, the monosyllabic answer is “no,” but that’s not entirely honest. If you give me a minute, I’ll give you an honest minute-long answer—maybe even a minute and a half. I’ve never been close to giving up, but as a human being, of course, I work very hard like many of my colleagues. We have a lot of pressure, and it happens that I wake up in the morning tired and think, “To hell with this.” Yes. And I haven’t given up and do not plan to give up for as far into the future as I can imagine. Although, I don’t know about 20 years from now; who can tell? But I have no plans, no considerations at the moment, because there is a very strong sense of responsibility—first and foremost to my colleagues, and also to our investors and creditors, and to the 11 million paying customers, not to mention the 300 million active users. Anyway, I would like us to keep trying to create products that work well. However, the strongest sense of responsibility is towards my colleagues, especially since practically everyone has invested a significant part of their wealth and their earnings in company stock. If we were to fail, which I don’t believe will happen, I want to have no regrets; I want to have given my all so that I can say there was nothing more we could have done. Also, I’m very competitive, and I really like the idea of being part of building what we hope will be the most valuable Italian company in history. I believe we will make it; we’ll see. Let’s talk about it again in three or four years. I wouldn’t leave; it would be like leaving a team that is aiming to win the World Cup and has a real chance—no, you can’t; you have to give it a shot. Even if it’s exhausting; actually, the closer you get to that goal, the harder it becomes. But at the same time, it’s also increasingly stimulating. So, no, I haven’t been there; I’ve obviously had many moments of discouragement and non-clinical depression; you understand, more a sense of disheartenment. However, I’ve never seriously considered giving up, and I am not contemplating it at the moment. Good. One of our co-founders left in 2018, so it can obviously happen. Um, okay, Luca, I’m really happy—or rather LF, as I’ve learned you’re called here. Yes, yes, they shorten everything. So, as I said, I’m really happy and I thank you for your time and the attention you dedicated to your answers, which I believe is worth much more than just time. I hope you all appreciated it, and who knows, maybe there will be a part two. We’ll see. See you soon. Thank you.

Simple App metrics.

According to company’s CEO comments in techcrunch over this funding round Simple app currently has:

• $160 million in ARR

• 700,000 subscribers (i.e. 228$ per subscriber)

• the company previously raised $10M before the current $35 million Series B

source: https://techcrunch.com/2025/10/01/kevin-harts-vc-firm-leads-35m-series-b-for-weight-loss-app-simple/

Interestingly, if we look at the sensortower data the app is doing around $260k in gross monthly revenue, or just around $3M in annual revenue. If the CEO’s claim of $160M revenue is accurate, this means that most (like 98% ) of the revenue goes through Stripe / outside of the app store. Using web funnels with stripe or other web acquiring option become a trend in recent years, as it allows to skip Apple’s rediculous 30% commission and provides better attribution data. But 98% from web funnel for a mobile-first product is quite surprising.

Checking the data further, the company’s website has a blog post announcing the raise. But in this post, there is no mention of 160M revenue. The post claims $100M revenue in 2024 and 800k subscribers. This gives a more reasonable $125 revenue per subscriber.

Source: https://simple.life/blog/raise-announcement/

So something is not adding up. Another weird thing, the number of installs in sensortower is very low for a mobile app with $100M in revenue. So there is some magic going on. It will be interesting to keep an eye on the company and its future announcements.

Condensed Summary

This discussion highlights why AI evaluations (evals) are essential for product managers building AI-driven features. It covers four eval types—code-based, human, LLM-as-judge, and user—and walks through a real-world example: creating a customer support chatbot for “On” running shoes. Key steps include drafting prompts with policy and product context, building a human-labeled golden dataset with rubrics, iteratively refining prompts, automating evals with LLMs, measuring match rates, and planning a scaled rollout with A/B tests and user feedback.

Detailed Bullet Points

📊 Types of AI Evaluations

• code-based eval: pass/fail checks via scripts (e.g., blocking competitor names)

• human eval: subject matter experts or PMs grade responses (thumbs up/down)

• LLM-as-judge eval: have an LLM label outputs at scale like a human judge

• user eval: gather downstream customer feedback as a business metric

🔧 Practical Example: “On” Running Shoes Agent

• goal: build a customer support bot for On running shoes on their website

• platform: use Anthropic Workbench to generate and refine the system prompt

• context inputs: user question, product information (shoe names, specs), return policy

🛠 Prompt Development & Iteration

• initial prompt: describe role, include variables for question, product info, policy

• use generate feature to auto-format placeholders and example dialogues

• test with a real question: “How do I return my On Cloud shoes after two months?”

• critique output: tone too formal, missing next-step details, may lack clarity on contacting support

📄 Creating a Human-Labeled Golden Dataset

• define evaluation rubric columns (product knowledge, policy compliance, tone)

• set criteria for good/average/bad in each dimension

• populate spreadsheet with ~5–10 example interactions to start

• debate labels among team members to align definitions and edge cases

• document notes alongside bad examples (e.g., LLM math errors on time windows)

🔄 Iterative Evaluation Process

• refine prompts based on rubric insights (e.g., inject friendliness, concise tone)

• use few-shot examples or system-level instructions to adjust style and rules

• test revised prompts on initial examples, re-grade against rubric

• repeat iteration loops until examples meet quality thresholds

🤖 Automating with LLM-as-Judge & Match Rate

• upload golden dataset into an evaluation tool (e.g., Arise platform)

• configure eval prompts that mirror rubric criteria for automated labeling

• add explanation requests so the LLM justifies each rating

• run multiple examples and models (Claude, GPT) to compare outputs

• calculate match rate: percentage of LLM labels matching human labels

• identify dimensions with low match rates (e.g., tone agreed only 1/5 times)

🚀 Scaling & Deployment Strategy

• internal testing phase: start with ~10 labeled examples for rapid iteration

• pre-production eval: expand to ~100 examples to boost statistical confidence

• A/B testing: route small % of live traffic through the bot (1–10%)

• dogfood: have employees use the bot to spot issues before wider launch

• user eval considerations: collect thumbs-up/down but interpret noise carefully—tie-break with deeper review

📝 Conclusions

• AI evals are foundational for delivering reliable AI products and mitigating hallucinations

• Four eval types cover technical, human, automated, and real-user feedback

• A structured loop—prompt draft, human labeling, rubric debate, automated eval, prompt refinement—drives quality

• Matching LLM-as-judge labels with human judgments ensures scalable, trustworthy assessments

• Phased rollout (internal → A/B → full) and continuous feedback help maintain performance and user satisfaction

Link to the video:

https://youtu.be/TL527yTpxlk?si=zR3CStdtZWQ5xsiZ

Life360 is a family safety and location-sharing mobile app that monetizes primarily through subscription memberships. Operating on a freemium model, it offers a free tier to attract a large user base and tiered premium plans (Silver, Gold, Platinum) for subscribers seeking extra features. This strategy has enabled Life360 to build a massive user base – over 79 million monthly active users (MAU) as of late 2024 – and steadily convert a portion of them into paying customers. The company’s rapid subscriber growth, rising average revenue per user, and strong retention have drawn investor attention, propelling Life360’s valuation into the multibillion-dollar range.

Subscriber Growth Over the Years

Life360’s paying subscriber count (measured in “paying circles,” essentially paid family groups) has surged over the past several years. In 2018, Life360 had only about 370,000 paying circles globally, and by mid-2024 this had grown to roughly 2 million. This exponential growth reflects Life360’s successful expansion in its core markets, especially the United States. In the U.S. alone, paid circles increased from approximately 280,000 in 2018 to 1.5 million by 2024, while international subscribers grew from about 93,000 to 562,000 in the same period. The momentum continued through 2023 and into 2024:

• 2021: ~1.2 million paying circles (post-IPO growth phase).

• 2022: 1.5 million paying circles globally (up 23% YoY).

• 2023: 1.8 million paying circles (up 21% YoY).

• Q4 2024: 2.3 million paying circles (25% YoY growth).

• Q1 2025: 2.4 million paying circles (another 26% YoY increase).

This consistent growth underscores the effectiveness of Life360’s model: the app’s large free user base provides a funnel for new subscribers, with steady conversion rates over time. In fact, Life360 notes that its freemium approach “attracts a large base of free users and converts them steadily into paying subscribers”. The company has also expanded through acquisitions (e.g. Tile trackers and Jiobit wearables) and new features, which help entice users into premium plans. Importantly, subscriber retention is high – net retention has been close to 100% in recent cohorts – ensuring that growth in subscribers translates into sustained revenue.

Average Revenue per Subscriber (ARPU)

Alongside user growth, Life360 has significantly increased its revenue per subscriber through pricing strategy and tiered offerings. The company’s premium plans range from about $7.99 up to $24.99 per month for U.S. users, and most subscribers opt for the mid-tier “Gold” plan (which accounts for ~81% of U.S. subscriptions, versus 15% on Silver and 4% on the top-tier Platinum). Over the years, Life360 has implemented price increases and introduced higher-value tiers, leading to a sharp rise in ARPU:

• In 2018, the average revenue per paying circle was only about $59 per year (in the U.S.), reflecting the older, lower-priced plans.

• By 2024, average revenue per paying circle had climbed to roughly $152 per year in the U.S., nearly triple the 2018 level. Globally, the blended ARPU is slightly lower due to international markets, but still around $128–$133 annual per subscriber (≈$11 per month) as of 2024.

This rising ARPU (approximately 6–8% YoY increase in recent quarters) has been driven by multiple factors: legacy subscribers moving to higher tiers, new subscribers opting for premium plans, and specific price hikes in key markets. For example, a major U.S. price increase in late 2022 boosted U.S. ARPU by 24% year-on-year in Q4 2023. Internationally, ARPU lags behind the U.S. (about $56 annually in 2024 on average for non-U.S. subs), but Life360 began rolling out multi-tier memberships and price updates overseas as well. In Q4 2024, international ARPU jumped 42% YoY after the introduction of new tiered plans in the UK, Australia, and New Zealand.

Overall, Life360’s global ARPU is now in the ~$130 per subscriber per year range, reflecting the company’s success in monetizing each user more effectively over time. This metric is crucial for a subscription app’s economics: higher ARPU means that each subscriber contributes more revenue, which, combined with subscriber growth, yields accelerating revenue gains.

Revenue Growth and Financial Performance

The combination of rapidly growing subscribers and rising ARPU has translated into impressive revenue growth. In full-year 2024, Life360’s total revenue reached $371 million, up 22% from the prior year. Subscription (recurring) revenue accounted for the bulk of this, about $278 million in 2024 (26% YoY growth), while smaller streams include hardware sales (Tile devices) and partnership/advertising revenue. Notably, Q4 2024 subscription revenue jumped 36% year-on-year, driven by the 25% increase in paying subscribers and ~6% higher ARPU that quarter. By Q1 2025, the momentum continued with total quarterly revenue up 32% YoY, as subscriber additions remained strong.

Life360’s operating metrics highlight a robust subscription business model: gross margins on subscription revenue exceed 85%, and as the user base scales, the company has started achieving positive cash flow and EBITDA. In 2024, Life360 delivered its first full year of positive adjusted EBITDA ($45.5M) and even a modest net profit in Q4. This improvement in profitability, alongside continued growth, has increased investor confidence. The company has also launched new monetization avenues (like in-app advertising in late 2024) to supplement subscription income. Management targets an additional >$1 per MAU in annual ad revenue within a few years, which could further boost average revenue per user beyond subscription fees.

Market Valuation and Subscribers

Life360’s public market valuation has risen dramatically as the company’s fundamentals strengthened. As of mid-2025, Life360’s market capitalization is around $5–6 billion USD.

How does this valuation correspond to the subscriber base? With approximately 2.4 million paying subscribers in early 2025, the market is effectively valuing Life360 at about $2,400 per subscriber (enterprise value divided by number of subs). Another way to view it: the stock’s valuation is roughly 18–19 times the annual revenue per subscriber (since ARPU is ~$130 and EV per sub is ~$2,400). These ratios indicate that investors are pricing in substantial future growth – they expect Life360 to continue expanding its subscriber count, increasing ARPU, or monetizing users via new streams (like ads or additional services) to eventually “grow into” this valuation. It’s worth noting that Life360’s valuation also equates to about $70 per monthly active user when considering all 80+ million MAUs (most of whom are free users), which underscores the potential value if even a fraction of those free users convert to paid plans or generate advertising revenue.

For context, comparing Life360’s value-per-subscriber to some other subscription-based tech companies highlights the premium attached to Life360’s growth. For example, Netflix – with a far larger base of ~300 million subscribers – has a market cap near $493 billion, which comes out to roughly $1,600 per subscriber. Music streamer Spotify, at ~$129 billion market value and 276 million premium subscribers, is about $470 per subscriber. Life360’s ~$2,400 per subscriber valuation is significantly higher than these mature businesses. However, this disparity reflects different business profiles: Life360’s subscribers generate a similar ARPU to Netflix (around $10–12/month), but Life360 is much smaller and growing faster (30–40% revenue growth vs. ~10% for Netflix/Spotify). Investors are effectively paying a higher multiple for Life360’s future expansion potential. The company’s goal of reaching 150 million MAUs and $1 billion in annual revenue in coming years indicates the scale of growth envisioned. If Life360 achieves those targets, the per-subscriber valuation would normalize as numbers scale.

Conclusion

Life360’s market cap of around $6 billion. The company’s challenge (and opportunity) will be to meet the high expectations: sustaining strong subscriber additions, growing ARPU (including through new features and advertising), and warding off competition from tech giants offering free location tools.

In summary, Life360 stands as a compelling case study in mobile app subscription monetization. Its valuation reflects a blend of current performance and future promise – the result of rapid subscriber expansion, effective monetizing of each user, and a clear vision to broaden its services. For observers of the mobile subscription economy, Life360 shows how converting a small fraction of a large user base into paying customers can translate into significant revenue and how, in turn, those subscribers can underpin a multi-billion dollar enterprise value in the public markets.

Sources: Life360 company filings and press releases; industry analyses and news reports; market data on comparable companies.

Here are the basic metrics for both companies side by side:

The crazy part is that Duolingo has 17 times higher valuation with 30% lower revenue at the same time. Duolingo trades at about 20 times its annual revenue, whereas Bumble trades at under 1 times revenue. The market’s valuation of each company has diverged dramatically. As of mid-2025, Duolingo’s market capitalization is about $17 billion, while Bumble’s is around $1 billion.

Why is Duolingo worth so much more per dollar of revenue?

A few key factors help explain this valuation gap:

• Growth and Future Potential: Duolingo’s robust growth (40%+ annually) signals a long runway of expansion – investors expect years of double-digit growth ahead. Additionally, Duolingo’s huge base of 500+ million registered users (with ~100M active) gives it a large funnel to convert more paying subscribers in the future. Bumble’s growth has stagnated, so the market isn’t projecting big future gains for Bumble’s top line. Essentially, Duolingo’s revenue might double in a couple of years if trends continue, whereas Bumble’s might barely budge – so investors assign Duolingo a rich “growth premium.”

• Market Narrative & Sector: Duolingo operates in the edtech/online learning space, which currently enjoys positive sentiment. The company has a compelling story of using gamification and AI in education, and it has branched into new learning verticals (music, math, etc.), suggesting it can expand its addressable market beyond just language learning. Bumble, in the dating app sector, faces a more skeptical market narrative – dating apps are seen as a tougher, more saturated market with slower growth and heavy competition (e.g. Match Group’s apps). There’s also perhaps less excitement or “hype” around dating tech compared to AI-driven education tech.

• Profitability and Strategy: Neither company is highly profitable at the moment (both have been near break-even on a net income basis). However, Bumble has a solid Adjusted EBITDA margin (~25+), indicating it is running a profitable core business but just isn’t growing much.

So back to the initial question. Is Bumble a buy? Probably yes. One of the reasons it’s just super undervalued and they are doing buybacks. Facing a slumping share price and investor pressure, Bumble’s management has taken steps to improve its financial metrics and shareholder value. Bumble authorized a large share repurchase program up to $450 million. In the fourth quarter of 2024 alone, Bumble repurchased about $40.3 million worth of shares, on top of ~$89 million repurchased in Q3. This is a significant amount for a company with a ~$1B market cap. Another reason, Bumble has also embarked on major cost-cutting measures. In June 2025, the company announced a restructuring with layoffs affecting 30% of its workforce. Roughly 240 employees were let go. This move is expected to save about $40 million annually in operating expenses.

• Harvey (San Francisco)

  • Funding: $300 m Series D — 12 Feb 2025, led by Sequoia, Coatue, Kleiner Perkins & GV

  • What it does: Full-stack Gen-AI workbench that BigLaw uses for contract analysis, diligence and litigation strategy.

• EvenUp (San Francisco)

  • Funding: $135 m Series D — 8 Oct 2024, led by Bain Capital Ventures with Premji Invest, Lightspeed, Bessemer & others

  • What it does: Personal-injury claims-intelligence platform that drafts, values and reviews case files automatically.

• Legora (Stockholm)

  • Funding: $80 m Series B — 21 May 2025, led by ICONIQ Growth & General Catalyst

  • What it does: Gen-AI copilot (Word/Outlook add-ins) that surfaces firm knowledge, drafts and redlines documents.

• Luminance (London/Cambridge)

  • Funding: $75 m Series C — 18 Feb 2025, led by Point72 Private Investments

  • What it does: Proprietary “Legal-Grade™” transformer that negotiates, red-flags and analyses contracts end-to-end.

• Supio (Seattle)

  • Funding: $60 m Series B — 30 Apr 2025, led by Sapphire Ventures with Mayfield & Thomson Reuters Ventures

  • What it does: Gen-AI document-intelligence stack tailored to plaintiff-side mass-tort and personal-injury firms.

• Lawhive (London)

  • Funding: $40 m Series A — 4 Dec 2024, co-led by GV (Google Ventures) & TQ Ventures

  • What it does: “AI lawyer” platform that automates client intake, casework and billing for small/consumer-law practices.

• Wordsmith AI (Edinburgh)

  • Funding: $25 m Series A — 3 Jun 2025, led by Index Ventures

  • What it does: Fleet-of-agents platform that answers business-unit queries and automates in-house legal workflows.

• Genie AI (London)

  • Funding: $17.8 m Series A — 23 Oct 2024, led by GV with Khosla Ventures

  • What it does: Open-source contract editor with 1,800+ templates and an agentic drafting/review assistant.

• Semeris (London)

  • Funding: $4.3 m Seed — 3 Feb 2025, led by Puma Growth Partners

  • What it does: AI parses 5,000+ CLO/ABS deals so finance lawyers can run “what’s-market” clause searches in seconds.

• Theo Ai (Palo Alto)

  • Funding: $4.2 m Seed — 19 May 2025, co-led by NextView Ventures & Collide Capital

  • What it does: Predicts litigation outcomes and settlement ranges, helping firms quantify case risk early.

So the total funding raised by AI legal startups is almost a billion dollars in just last 12 months. All of them are in B2B segment, i.e. providing services to either different types of legal teams. But what about B2C case? Direct to consumer legal AI? Seems like a huge opportunity and some one should eventually go for it.

When people talk about mobile subscription apps they usually think of fitness trackers, meditation or note-taking apps. A global fintech is hardly the first thing that comes to mind. Yet Revolut, originally started as a low-fee FX card, now pulls in about half a billion dollars a year from subscriptions alone, up 74 % year-over-year.

Revolut’s 2024 numbers:

• Currency Exchange: $0.5Bn (+60%)

• Subscriptions: $0.5Bn (+74%)

• Invest & Crypto: $0.6Bn (+298%)

• Card Payments: $0.9Bn (+43%)

• Interest Income: $1.0Bn (+58%)

So Revolut’s revenue from subscriptions is the same as on currency exchange and growing even faster. Subscriptions now contribute ≈12 % of total revenue.

Revolut offers several tiers of subscription. Here’s what you get at different tiers (listing just some of the perks here):

• Standard - FreeBasic account, FX at 1 % spread

• Plus £3.99/mo - Purchase protection, disposable cards

• Premium£7.99 - Travel insurance, higher FX limits

• Metal£14.99 - Cashback, concierge, higher ATM limits

• Ultra £45 - 3×WeWork day-passes, airport lounge access, travel e-sim, NordVPN, 20 ClassPass credits, free weekend FX, and more.

A single day WeWork pass in London costs ~£40 on its own, so three passes already exceed the £45 fee. Add a £12/month NordVPN subscription, airport lounge access, and travel e-sim and the tier feels like a no-brainer for any frequent travellers.

Sizing the paid subscribers base (back-of-the-envelope)

Revolut had around 52M users in 2024 with subscription revenue of $0.5 bn. Let’s assume the weighted avg. price ≈ $15/mo → $180/yr

→ Paying users ≈ 2.8 m, or ~5 % of Revolut’s 52.5 m customers.

In other words, a single-digit share of the user base delivers an outsized chunk of revenue and plenty of upside remains as Revolut upsells more free customers. Looks like subscription side of the business has potential to become the major source of Revolut’s revenue and its growth in the coming years.

And unlike regular apps on the App Store Revolut does not have to pay 30% to Apple and 5-10% VAT on every sale. Amazing business model, isn’t it?

Sources: Revolut 2024 annual report (published 24 Apr 2025); LinkedIn post by Dmitry Zlokazov, Head of Product at Revolut

Runna was founded less than 4 years ago in 2021. It’s an app specializing in personalized running training plans and coaching. Users input their running experience and goals and the app generates a customized training program that adapts to their progress​.

Runna monetizes with subscription (around $120 per year). The company has raised a total of about $10M in venture funding. It first raised £485,000 via a crowdfunding campaign in 2021, attracting early support from hundreds of its initial users. This was followed by a £2.25M seed round led by Eka Ventures. In 2023 Runna raised a £5 million ($6.3M) Series A round.

Strava is a leading fitness social network app founded in 2009. Over the years it has grown into one of the world’s largest online communities for athletes. Strava’s monetizes with subscription (about $80 per year). Strava’s revenue estimated at $270M in 2023 by Bloomberg. Strava rasied around $150M in total funding

Strava announced that it acquired Runna on April 17, 2025.

Strategic Rationale

Strava’s motivation for acquiring Runna centers on filling a gap in its own product offerings. Despite Strava’s popularity among runners, the platform lacks built-in training plans or coaching. In Strava’s research, users identified guidance and structured training as a key missing piece of the experience​. For runners who wanted a plan to prepare for their next race or improve their performance, Strava offered tracking and social motivation, but not coaching. Many users need more than just to record and share runs, they want to know what workouts to do next. Runna was purpose-built to provide that guidance.

By acquiring Runna, Strava instantly closes that gap with a proven product​. Runna brings to Strava a library of training programs, an experienced coaching team, and a user base of runners. It would have taken Strava significant time and resources to build a comparable training platform from scratch. Strava gains Runna’s technology and talent. The likely game plan, as hinted by Strava’s leadership, is to integrate Runna’s coaching seamlessly into the Strava ecosystem but without alienating Runna’s existing users. Initially, Strava plans to operate Runna as an independent app.

Runna Aquisition Price

The acquisition price was not disclosed. App’s revenue in March 2025 was around $3M per month on iOS and $400k on Android per Sensortower. So roughly $3.5M per month in total or around $40M in annual revenue run rate (running app’s run rate :).

Strava’s monthly revenue was about $9M on iOS and $3M on Android at the same time, so roughly $150M per year.

If I had to guess I’d say Runna’s valuation was at least $100M and potentially more. Let’s assume a modest 3x revenue multiple, that means $120M valuation. The actual valuation usually depends on the growth rate as well. So if Runna was growing quickly the valuation might have been even higher, and if the revenue did not grow that much in recent months then probably the valuation could have been in the range of $80M-$100M.

It’s worth saying that all of these companies are making over $100M a year, so they’re definitely on the higher end. Smaller companies tend to get lower valuation multiples, both when they raise money and when they get acquired.

2024 Consumer App Metrics with Valuation Multiples

All of the companies below are mobile first and monetizes with subscription. Data on some of them is readily available as they are publicly traded and disclose detailed financial information and metrics, for others data from different sources related to their fundraising anouncements and other sources. Overall there are 7 companies in this list: Duolingo, Life360, Bumble, Flo, Headspace, Calm, Strava.

* Annualized based on Q3 2024 revenue of $192.6M
** Annualized based on Q4 2024 revenue of $261.6M

The median EV/Revenue multiple is around 5x, while the average is closer to 8x, pulled up by outliers like Duolingo. This gives a good sense of how top companies in the mobile subscription space are being valued. The best performers—those with strong brand, growth, and retention—can command much higher multiples, while most others sit in the 3–6x range.

Investors in Subscription Apps:

Summary Table of Investors

And here’s a condensed list of VCs and key investors in these companies. Because maybe… just maybe… you were looking for the list of investors who invest in mobile apps?

Duolingo: Union Square Ventures, NEA, Kleiner Perkins, CapitalG, General Atlantic, Drive Capital

Life360: Bessemer Venture Partners, DCM Ventures, Fontinalis Partners, Allianz X, Bryant Stibel

Bumble: Blackstone, Accel, Greycroft, Bain Capital Ventures

Flo: VNV Global, Mangrove Capital Partners, General Atlantic, Target Global, EQT Ventures

Headspace: The Chernin Group, Spectrum Equity, Advance Venture Partners, Blisce/, Insight Partners

Calm: Insight Partners, Lightspeed Venture Partners, TPG Growth, Goldman Sachs, Marc Benioff

Strava: Sequoia Capital, Madrone Capital Partners, TCV, Sigma Partners, Dragoneer Investment Group

Note that two investors had invested in two companies from the list:

• General Atlantic: Appears in Duolingo and Flo, two companies with different focuses (language learning and women’s health) but similar subscription-driven, mobile-first models. This suggests General Atlantic targets high-growth consumer apps with scalable monetization.

• Insight Partners: Invested in Headspace and Calm, direct competitors in the meditation and mental health market. This overlap indicates a strategic play in the booming wellness sector, possibly diversifying bets across leading players.

I have an online business where customers can contact us via email. The questions are usually about pricing, product features, bugs, or payment issues. They are not complex, but to answer you do need to know the details of our product. You also need to understand a user's specific case. The requests are in different languages.

My solution uses 3 platforms: OpenAI, Gmail, and Zappier to stitch them together.

When I got tired of repetitive support work and thought of hiring our first support agent it hit me that there should be a better way. The support requests are somewhat typical and my guess was chatGPT-4 should be able to handle them if given the right context upfront.

I checked some SaaS solutions available on the market. The one that might work is Intercom. But the pricing is crazy. They want $1 per "resolution", i.e. a dollar per chat. No, thank you Intercom.

So I decided to build it myself. It appeared to be easy, with no coding required. You can do the same for your business in an hour.

Here are the exact steps.

Part 1 - create an Assistant in OpenAI.

1. Sign up for an OpenAI API account.

2. Navigate to the Playground and select Assistant.

   

3. Create an assistant. That's where the magic happens. We need the AI to know the details of our business. The best replies to the standard questions, frequent issues, details of the pricing plans, etc.

4. First, type in the prompt with basic instructions like "You are a support agent named Anna for our company providing so and so service. Please be polite and concise, and always greet users.” Ask to reply in the same language the question is if you have international clients.

   

5. Then provide the specific details on how to handle standard requests. You can provide instructions as plain text in the same prompt field. But you can also attach them as a text or PDF file. Using a file can be easier so you can keep a copy of the file that you update with new information and product changes, and update this file from time to time in OpenAI interface.

6. Test the assistant. Go and ask the questions in that same interface. Just copy the last 5-10 questions you got from your customers and see if your assistant provides the expected replies or if you need to tweak your instructions.

Part 2 - connect your Assistant to gmail using Zappier and setup the automation

1. Create a Zappier account and click "Create" to create new automation.

2. You will need to create only free steps in that automation. The first step is to launch automation based on new inbound email. Chose Gmail, connect your account, and select "new email" as event triggering the automation.

   

3. The second step in the automation is ChatGPT - a conversation with an assistant. Connect ChatGPT, the event is: Conversation With Assistant. Provide the required login details. In the configuration of this task, you should use the "body plain" field from the previous step's email in the message field. You must select your assistant from the list of assistants.

4. The third step is again Gmail. The event is "create draft reply" if you want to review the replies manually. But you can also set it to send emails automatically. In the configuration of this event, you need to select the thread ID from the first step in the Thread field. You need to select the email address from the first email in the field "To". And you add the "from" field. At the end, you add the response from the ChatGPT step as the body of the email.

That's it. You are ready to go. Test the automation and see how it works. You can add additional steps like automatically classifying each inbound request to collect statistics and understand better what needs to be done to reduce inbound inquiries. This can be done as an additional step that sends the message to ChatGPT to add the category to the request and then save it into Google Sheets where you can gather the statistics.

The whole thing cost me around $50/month to handle around 1000 monthly inbound support emails. The cost is mostly for Zappier, it's cheap to handle that number of short emails with ChatGPT. So, kids, build your own AI agents, and don't get your business bankrupt by Intercom.

If you have any questions you can reach me at @mstysin on X.

I was interested in comparing the valuation of Flo and Duolingo - both consumer apps monetizing with subscriptions on the app stores at a significant scale. I picked some numbers from the company's releases, Techcrunch articles, and the company's CEO posts on Linkedin. I've added the data from Duolingo's latest quarterly report. Some numbers are rounded or estimated, so might be slightly off. However, the purpose was to provide a high-level comparison of the growth rate, scale, and valuation multiple.

Flo had around 68 million MAU (monthly active users) at the time of the raise, up from 54 million a year earlier. So the user growth rate is ±25%

The number of paid subscribers is 5 million. Revenue estimated at $200M.

Here is the side-by-side comparison of the main metrics, revenue, growth, and valuation. Flo is valued at 5x revenue, while Duolingo is valued at a whopping 17x revenue.

Why the difference? There could be several explanations. Overall, a 17x revenue multiple for any business is high and raises concerns about whether it can be sustained, and if the company can deliver the expected growth. One bad quarterly report could send shares down and lead to a significant adjustment in the valuation.

Duolingo Monetization Model

App Store is full of apps in categories from productivity to education monetizing with subscriptions. There are a number of apps with gross revenue in the range of $1 to $10 million a year that are sold for 2-3x revenue. The multiple is significantly lower compared to SaaS companies that usually valued at 6 to 10 revenue. The reason for a lower multiple is pretty simple, consumer apps have a much higher churn rate with 50%+ annual churn being normal. SaaS companies on the other hand usually have high retention and sometimes negative churn due to revenue expansion from existing clients. On top of it SaaS companies can be highly profitable, while consumer apps tend to be driven by performance marketing that takes a significant chunk of the revenue.

Now let's have a look at Duolingo's revenue. At the IPO it was at around $200 million. Trailing twelve months revenue as of June 2024 is $580 million. So if we apply a revenue multiple typical to smaller consumer apps, the market cap at the time of IPO would be just around $500 million and around $1.5B now. But Duolingo's market cap was $3.6 and currently trades at a whopping $8.5 billion. This implies a 13x revenue multiple. The multiple many SaaS companies would be jealous of.

The next logical step would be to look at the company's profitability and growth rates. Maybe we'll find the explanation for this extraordinarily high valuation there.

The company has been slightly loss-making in the last four years. The revenue has been growing by around 50% per year in the last 4 years with just a slight slowdown of 43% last year.

So not profitable consumer app is valued at 13x revenue. I can't help but notice that the company is probably overvalued by 3-4x. Applying the 3x revenue multiple it could be valued at $1.5B now.

Below are some interesting data points from Duolingo's annual report that might be interesting for those who are building mobile apps with similar monetization. The company tracks and discloses the following operational metrics: MAUs, DAUs, paid subscribers, subscription bookings, total bookings. 6 million subscribers were paying for Duolingo at the end of 2023, up from 4.2 just a year earlier.

As of December 31, 2023, subscribers made up 8.3% of MAUs (monthly active users) as compared to 7.8% during 2022.

Total bookings include subscription bookings, income from advertising networks for advertisements served to our users, purchases of the Duolingo English Test, and in-app purchases of virtual goods. So the company earned more than $120 million in 2023 on top of revenue from subscriptions.

While the company is growing and does look like a good business all in all it feels very overvalued. Non the less, the company is a stunning example of a veru succesful consumer app built on App Store. If you are in that industry it’s good to have an eye on it, it’s metrics, and learn from it.

Amplitude is the leading analytics solution for mobile-first and cross-platform companies. Anyone building mobile apps knows Amplitude as the main solution to track any in-app events and metrics from MAU and DAU to any custom conversions and retention metrics. It's powerful yet simple-to-use analytics that allows the creation of flexible dashboards and charts to understand users and make business decisions.

Amplitude was founded in 2014, a time of rapid growth for mobile apps. The company IPOed in 2021 at a $5B valuation with revenue of around $140M, growing 50% year over year. The revenue multiple was a staggering 35x. The shares of the company climbed from $50 to $80 in the months following the IPO, pushing the valuation even further to $8B. What was the revenue multiple then? An insane 60x revenue.

Within just four months of the IPO, the shares fell below the IPO level, and then the fall accelerated. Three years later, the company's shares traded at $8, with the company valued at roughly $1B. Meanwhile, the revenue grew to $280M. So the current multiple is just about 3x revenue.

Slow revenue growth is one of the reasons for the current low valuation, besides the SaaS bubble valuations of 2021. The annual growth rate is below 10%. The company is also loss-making, with almost $100M in annual loss. Selling and Marketing expenses are the largest expense category, around 55% of revenue. So if the company wants to get closer to profitability, it could reduce these expenses, meaning potential redundancies in the sales team. According to LinkedIn, the company's headcount is just below 900, with 150 in sales, and it has been steady for the last two years.

I've used Amplitude for several of my projects, and the product is great. The company is also considered one of the most successful companies out of YC, and obviously, this was a successful journey for the founders. Yet investing in the company's shares at IPO could have turned into a massive loss. So great products do not always mean great returns to investors. With the current growth rate, it's hard to expect the company to get anywhere near the valuation it had at the IPO years ago.