Chapter 4 of 15 · Business Models and Revenue
1 The bottom line
Sell a subscription with a metered credit allowance, then let heavy users buy more through consumables. This covers real AI costs and lets your biggest spenders pay what the value is worth.
2 Key takeaways
- Cap what the subscription includes. AI has a real cost per use, and flat all-you-can-eat pricing rules out costly features.
- OpenArt removed unlimited credits because video would have bankrupted it
Coco Mao, OpenArt26:00. - StackBlitz users burned through a $9 plan in 48 hours

Eric Simons, Bolt31:51.
- OpenArt removed unlimited credits because video would have bankrupted it
- Let anyone who hits the cap buy more. Spend has no ceiling, so the top slice pays for the value it gets.
- Sway's credits paywall brings about 11% of revenue, and photo packs are the biggest upsell

Daniel Heintzman, Sway AI20:15. - OpenArt users complained loudly about the new limits but kept buying add-ons
Coco Mao, OpenArt26:00.
- Sway's credits paywall brings about 11% of revenue, and photo packs are the biggest upsell
- Name the unit in words your audience already uses. Creators think in credits, but consumers do not want to think about tokens.
- The ElevenLabs creator app has five credit tiers, while ElevenReader has two plans

Jack McDermott, ElevenLabs34:20. - Switching ElevenReader to hours of listening was the biggest win of the year

Jack McDermott and Tanmay Jain, ElevenLabs54:30.
- The ElevenLabs creator app has five credit tiers, while ElevenReader has two plans
- Design consumables to amplify the subscription, not replace it. Boosts add revenue without cannibalizing the tiers.
- Tinder built Boost and Super Like as force multipliers of subscription features

Ravi Mehta, Tinder24:07.
- Tinder built Boost and Super Like as force multipliers of subscription features
- Study your biggest spenders and expect to reprice often. Your guesses about whales are often wrong, and costs shift as capabilities change.
- Tinder assumed its top spenders were wealthy, but they were military, sales and newly relocated people
Ravi Mehta, Outpace36:02. - Tolan expects to change its pricing every few months

Ajay Mehta, Tolan42:08.
- Tinder assumed its top spenders were wealthy, but they were military, sales and newly relocated people
drop the floor and raise the ceiling
Rik Haandrikman, RevenueCat
2.1 Pick the model before the price
How you charge matters more than how much, and AI cost no longer follows time spent.
- Decide the model and the metric first, then the price.
Madhavan Ramanujam, Simon-Kucher1:00:10
- Choose the model (subscription, pay as you go, freemium), then the metric, then the price structure.
- Choose a metric customers see as fair and tied to value.
- Segment moved from charging per API to monthly tracked users, and Michelin sold a longer-lasting tire by charging per mile.
- Choose subscription or usage by customer situation, not fashion.
Madhavan Ramanujam, Simon-Kucher1:04:30
- Subscription fits predictable bills, steady usage, or a simpler story like Spotify and Netflix.
- Usage fits low-commitment onboarding, demand for fairness, or costs that scale with use.
- Only charge on usage if you can track the metric.
- Hybrids like HubSpot, a fixed base plus overage, also work.
- Assume background agent work breaks the link between time in the app and cost.

Ajay Mehta, Tolan42:08
- Early on, time in the app tracked cost. Agentic features no longer follow that rule.
- A tab that tracks Dodgers scores each morning costs money while the user is not in the app.
2.2 Build the subscription around a credit allowance
Give each tier a fixed amount of AI use, so what a subscriber can consume matches what you pay for.
- Give free users a small allowance and paid users a larger one.

Rik Haandrikman, RevenueCat31:02
- Let anyone buy more. The low floor lets many people pay a little.
- Define paywall tiers by credits per day.
Steve P. Young, Photo Revive AI3:15 describes a basic plan with 240 credits a day and a higher plan with 570. - Tolan cut its trial from 7 days to 3 to limit exposure to runaway AI costs.
- Replace unlimited with hard limits plus add-on packs.
Coco Mao, OpenArt26:00
- Unlimited credits worked for images but would have bankrupted OpenArt once video launched.
- Some users asked for higher add-on purchase limits, and OpenArt expanded them.
- StackBlitz had only a $9 plan and shipped upgrade tiers within a week, once users asked how to buy more. Competitors later copied the model

Eric Simons, Bolt31:51.
- Set a quota per tier and mix fixed and variable parts.
Naomi Ionita, Evernote47:12
- Offer good, better, best tiers, each with a quota on your value metric. Hitting the limit triggers an upgrade or overages.
- Only around 5 percent of SaaS companies use pure usage pricing, because CFOs want predictable budgets.
- A value metric typically doubles expansion revenue, but downgrades run about 20 to 25 percent higher
Patrick Campbell, ProfitWell21:00.
2.3 Name the unit your customer understands
Keep credits for people who already think in credits, and sell everyone else something they can picture.
- Use credits for prosumers and simple plans for consumers, even on the same costly API.

Jack McDermott, ElevenLabs34:20
- Start with what a consumer wants to subscribe to and work backwards.
- ElevenReader aligned cost to hours of listening. The creator app keeps credits because its users were trained on them

Jack McDermott and Tanmay Jain, ElevenLabs54:30. - Real AI cost of goods still exists, but it should not be how you talk to consumers.
- Charge on the unit of value so price rises with use.
Naomi Ionita, Evernote16:27
- Examples are API calls, messages sent, terabytes stored or words written.
- The metric also helps you understand and target different segments.
- Show users what each AI feature costs to run.

Eugenia Kuyda, Wabi45:05
- Users pay for what each mini app costs, like giving small pocket money to a little army of apps.
- Creators whose apps are remixed or heavily used get a cut.
2.4 Sell top-ups and consumables on top
Consumables raise the ceiling, so the users who value the product most can pay most.
- Let heavy users pay beyond the cap.
Phil Carter, Elemental Growth58:46
- Borrow B2B usage pricing: base, pro and max tiers plus paid credits for users who exceed the cap.
- If the most a customer can spend is $8 a month, you leave demand uncaptured.

Jacob Eiting, RevenueCat36:44 says developers tend to undervalue their own product. - Extra purchases for your most loyal users also keep room for free promotions
Steve P. Young, App Masters6:55. - Consumers now accept $20 to $200 a month plans that would have been rejected three years ago.
- 35% of consumer subscription apps now offer lifetime plans and/or in-app purchases.
- Leave consumable spend uncapped.

Daniel Heintzman, Sway AI20:15
- Sway sold AI photo packs at $14.99 for one and $49.99 for ten, on top of the subscription.
- A credits upsell for people who ran the flow repeatedly added more.
- Require a subscription before users can buy credits.
Coco Mao, OpenArt29:45
- This ties credit purchases to recurring revenue and habit, rather than credits bought and forgotten.
- Tolan's boosts let a user who hits a tier limit buy extra chat time without a higher tier. Ajay Mehta says they have not nailed one-off purchases yet

Tolan24:33.
- Make consumables multiply what the subscription gives.

Ravi Mehta, Tinder24:07
- A Gold user who has gone through everyone who liked them buys a Boost for more likes, or a Super Like for someone they really want.
- Evaluate how features interact, not just what each does alone.
- Recipients of a Super Like are three times more likely to swipe right

Ravi Mehta, Tinder23:04.
2.5 Find your biggest spenders and make them offers
Learn who your whales really are, then give each spend segment a next step.
- Talk to your biggest spenders before assuming why they pay.
Ravi Mehta, Outpace36:02
- A single-digit percentage of Tinder users drove a large share of a la carte revenue, spending hundreds of dollars a month.
- A usability study showed an intense use case, and these users compared the price to the cost of dating.
- This led to Tinder Platinum, a third tier with bundled consumables.
- Segment by spend and show better offers to move people up.
Steve P. Young and a guest consultant, App Masters5:54
- Split users into buyers and non-buyers, and by how much they have spent.
- Raise the offer for past buyers, for example consumables from $5 to $7, $8 or $9.
- Bundle two products and make one free with the other. The guest calls the results very amazing for users seeing the paywall for the first time
App Masters5:09.
- Plan a sequence of offers, not one.
Tom Hammond, TheoremReach8:48
- If they buy, offer something bigger. If they do not, follow with a different offer.
- In Slotomania, a dismissed 1.8 million coin offer was followed by a 2.25 million coin offer with a two minute timer.
- Show offers after a free gift, when players feel thankful
Tom Hammond, TheoremReach3:32. An offer at minute three can fail, but by minute eight players are more likely to convert
TheoremReach8:59.
2.6 Make buying more simple
Every extra step costs you purchases from the users most willing to pay.
- Charge upsells in one click with the saved web payment token.
Igor Lyubimov, Headway26:35
- After the first web payment, Stripe keeps the token, so users do not re-enter card details.
- Stack upsells at different prices so users do not think they were double charged, and add a lower-priced downsell page.
- Headway's upsells are priced at 14.59 and 9.59 and charged instantly on click.
- Charge inside the app with the card saved from the web funnel.
Igor Lyubimov, Headway29:21
- In the US, one API method does it with no web view, and you skip the 30% App Store fee.
- Headway shows another upsell when the app launches.
- Make upgrading after a limit hit smooth.
Phil Carter, Elemental Growth58:46
- Phil and David both upgraded Claude repeatedly after hitting usage limits, and David hit friction doing it.
2.7 Where they disagree
Credits, outcomes and light purchases each fit different products.
- Credits suit assisted tools, but outcome pricing suits autonomous agents.
Madhavan Ramanujam, Simon-Kucher42:08
- High attribution with a human in the loop fits seats plus credits, and Cursor sits there. High autonomy and high attribution fits outcome pricing.
Bret Taylor, Sierra1:07:42 says usage and value are not correlated in AI, so Sierra charges a pre-negotiated rate per resolution.
Eoghan McCabe, Intercom23:45 charges $0.99 per resolved ticket, against $20 to $30 for a human-resolved one.- Do not rush in without attribution. About 5% of companies are in true outcome pricing, and he expects 25% in three years
Madhavan Ramanujam45:40.
Brian Balfour, Reforge43:18 warns that competition erodes outcome pricing without a durable moat.
- Uncapped spend suits some products, but light purchases suit others.

Sway AI20:15 relies on there being no cap to spend.
Matthew Hall, Crossy Road14:26 keeps the game light on the wallet, with no $99 transactions and permanent purchases.
- There is still no settled AI pricing unit.

Krithika Shankarraman, OpenAI1:02:00
- Test whether customers see value per seat, in hours saved, or in new things they can do.
- For agents and code generators, the unit of completion is still unresolved.
Sources
- How Tolan 3X'd Subscriber Conversion after Shifting to Freemium ·
Subversive - How Wabi is Building “YouTube for Software” with a Platform for Creating Mini Apps ·
Subversive - I speedran a new app from 0 to $2m/year in 6 months (here's how) ·
Superwall - The AI Growth Playbook for Subscription Apps – Phil Carter, Elemental Growth ·
Sub Club - How ElevenLabs is Building Consumer Apps on its $11B Voice AI Platform ·
Subversive - Insights from RevenueCat’s 2026 State of Subscription Apps Report ·
Subversive - This AI Photo Editing App Makes $500K a Month ·
App Masters - How ElevenLabs Ships So Fast With Small Teams ·
Sub Club - Copy Headway’s $100M+ Web2App Funnel Strategy (That Converts Like Crazy) ·
App Masters - How OpenArt is Using AI to Reinvent Content Creation ·
Subversive - How Intercom rose from the ashes by betting everything on AI | Eoghan McCabe (founder and CEO) ·
Lenny's Podcast - Why ChatGPT will be the next big growth channel (and how to capitalize on it) | Brian Balfour ·
Lenny's Podcast - He saved OpenAI, invented the “Like” button, and built Google Maps: Bret Taylor (Sierra) ·
Lenny's Podcast - Pricing your AI product: Lessons from 400+ companies and 50 unicorns | Madhavan Ramanujam ·
Lenny's Podcast - Growth tactics from OpenAI and Stripe’s first marketer | Krithika Shankarraman ·
Lenny's Podcast - Insights from RevenueCat’s 2025 State of Subscription Apps Report ·
Subversive - Inside Bolt: From near-death to one of the fastest-growing products in history | Eric Simons ·
Lenny's Podcast - How Tinder's Monetization Strategy Propelled it to $1B+ in Revenue ·
Subversive - 2 Secret Retention & Monetization Strategies (Stolen From 7-Figure Apps) #retentionstrategies ·
App Masters - 10 lessons on bootstrapping a $200m business | Patrick Campbell (ProfitWell) ·
Lenny's Podcast - How to build your product strategy stack | Ravi Mehta (Tinder, Facebook, Tripadvisor, Outpace) ·
Lenny's Podcast - How to price your product | Naomi Ionita (Menlo Ventures) ·
Lenny's Podcast - The art and science of pricing | Madhavan Ramanujam (Monetizing Innovation, Simon-Kucher) ·
Lenny's Podcast - Auditing The Most Downloaded Game - Subway Surfers ·
App Masters - 5 Ways to Increase Mobile Game Revenue ·
App Masters - Crossy Road Creator Matthew Hall (Audio Masterclass) ·
App Masters - Hottest Monetization Trends for Mobile Games ·
App Masters - Live Q&A Call: Subscription Apps, Pricing Models & Pre-Launch Strategies ·
App Masters
Part of Business Models and Revenue in the growth library.