Guide D · Monetisation models
Decide how you charge
How you charge shapes the sales motion as much as the product does. This guide covers the models software companies use today, when each one fits, and what each asks of the motion you chose in module 01.
Where this sits in your plan
Georg Tacke, co-author of Monetizing Innovation with Madhavan Ramanujam, puts the order of decisions plainly. "How you charge is often much more important than how much you charge." So decide the model here, then set the price with Guide E. The examples come from startups and scale-ups.
Module 02 asks how pricing and packaging were set, and when you last tested them with buyers (m02-q9). Its Design rows ask for the chosen model with the reason for it. At Series B they ask for a model for each product.
The model has to agree with the motion you chose in module 01 (m01-q5 at Seed and Series A, m01-q8 at Series B). A product priced at a few pounds a month can't pay for a field sales team, and a large platform contract won't close through a card checkout. Brian Balfour calls this the fit between channel and model. At Series B, module 01 asks which of the four fits is weakest (NEW-01c), and two of the four involve the model.
A model the founders have chosen but sales and finance haven't agreed scores 2 on the toolkit's scale. Written and agreed earns a 3. A 4 needs evidence, such as a recent buyer test.
This guide decides what the customer pays for and how the bill is built. Guide E sets how much, and how you hold it. Choose the unit the bill scales on with the model, and test it with buyers using Guide E's method.
What changed in 2025 and 2026
Seats are no longer the default. In Kyle Poyar's 2026 survey of 230 B2B software and AI companies, hybrid pricing was the most common model at 37% of companies, and 29% already sold AI credits, with another 33% planning to. Only 5% of the investors surveyed preferred seat-based pricing. Three in four companies had changed their pricing or packaging in the previous year.
The pressure comes from AI. A seat doesn't track the cost of the compute behind it, and an agent does work that no seat holder does. So vendors are adding metered layers to seats, or replacing seats with outcomes.
The models in use
| Model | You charge for | It fits when | What it asks of GTM |
|---|---|---|---|
| Flat fee | One price for the account or plan, whatever the use | Use varies little between customers, and buyers want a quick yes | A price a buyer can accept without a call. Growth comes from moving customers to a higher plan, so the plan needs a reason to upgrade |
| Per seat | Each named user, usually billed annually | Every user gets value, and value grows with the customer's headcount | Land and expand by adding users. Expansion depends on the customer hiring |
| Usage | A countable unit, such as messages, compute, API calls or storage | Value tracks the unit, and buyers want to start small | Self-serve entry, metering and billing from the start, and a sales team that sells committed spend |
| Hybrid | A base fee or seats, plus metered usage or credits | You need a predictable base, but your costs or the customer's value scale with use | Clear packaging, spend caps or alerts, and a team that can explain the bill |
| Outcome | A result, such as a resolved support conversation or a qualified lead | The outcome can be counted, credited to you and agreed in advance | Outcome definitions in the contract, reporting the customer trusts, and a minimum or platform fee |
| Transaction | A share of the money that flows through you | You sit in payments, commerce, marketplaces or billing | Published rates that large customers negotiate down, and revenue that moves with your customers' sales |
The 'It fits when' and 'What it asks of GTM' columns are Ascent's summary. The examples below show each model in use, with prices as published on 4 November 2026.
| Model | Examples |
|---|---|
| Per seat | Linear charges $10 or $16 per user a month, billed yearly, with a free plan capped at 250 issues. Pipedrive sells per-user plans from US$14 to US$79 a month, billed annually. |
| Usage | Twilio charges per SMS segment sent in the UK. Neon moved its paid plans to fully usage-based pricing with no monthly minimum in August 2025. Snowflake describes itself as a consumption-based business in its annual report. |
| Hybrid | Supabase Pro starts at $25 a month with usage quotas, and its spend cap is on by default. Attio includes 500 credits a month with each seat. GitHub Copilot kept its seat prices and began metering AI use through credits from 1 June 2026. |
| Outcome | Intercom's Fin charges $0.99 per outcome, with a minimum of 50 outcomes a month on third-party helpdesks. Zendesk announced in August 2024 that customers would pay only for issues its AI resolves on its own. Sierra says it is paid only when it completes a task. |
| Transaction | Mollie charges 1.80% + €0.25 on EEA consumer cards. Paddle charges 5% + 50¢ per checkout transaction with no monthly fee. Shopify combines a subscription with a payments rate. |
Where each model goes wrong
- Flat fee and per seat. Companies using either named expansion revenue as their biggest challenge in the 2026 survey. Bessemer describes AI-native companies as abandoning seats.
- Usage. Customers optimise their spend when budgets tighten. Snowflake's annual report warns that unexpected changes in consumption can make results differ from its forecasts. Bessemer reports that some customers became wary of using a product priced this way.
- Hybrid. Companies using it were the happiest in the 2026 survey, though they also found it the hardest to explain.
- Outcome. In the 2025 survey, 5% of companies used it as their main model and 25% expected to by 2028. The same survey judged it out of reach for 95% of companies for now, because few can define an outcome that both sides can count, attribute and predict. The 2026 survey adds that enterprise buyers want spend they can forecast.
- Transaction. Revenue follows your customers' sales, so plan for a downturn in theirs.
Pricing AI features
Many vendors now sell AI use through credits. These sit on top of seats and can later be tied to outcomes. Salesforce Agentforce replaced its $2-per-conversation price with Flex Credits in May 2025, at $0.10 per action. Clay split its single credit into data credits and actions in March 2026, so customers could see what they paid for data and what they paid for the platform.
When Cursor moved its Pro plan to a pool of included usage in June 2025, it apologised publicly and offered refunds. Bessemer warns that companies that don't track the true cost of AI use early can grow into negative margins without noticing.
Ramanujam gives a quick test for whether outcome pricing is within reach. In a July 2025 conversation with Lenny Rachitsky, he places an AI product on two axes. Autonomy asks whether it does the work on its own. Attribution asks whether you can show the result came from you. Outcome pricing belongs with products that have "great autonomy and great attribution". If yours is weak on either axis, a seat or usage base with credits is the safer choice for this year's plan.
Free plans and trials are a way in
A free plan or trial sits on top of a paid model. Module 02's Series B Design row lists freemium among the models, so name both in the plan.
The Growth Unhinged free-to-paid report (February 2026) covered 200 B2B products. It found free trials were the main entry point for 57% and freemium for 26%. The median free-to-paid conversion was 8%. Trials that asked for a card converted at about 30%, more than five times the rate of trials that didn't.
Airtable uses a reverse trial. A full trial drops to the free plan when it ends, so users feel what they'd lose. Free users now cost more to serve, because AI features run up token bills. Decide what the free plan includes with that in mind.
The way in also feeds m02-q4 at Seed, and the ready-to-pay signals in module 03 (m03-q13 at Seed, m03-q14 at Series A and B).
Pressure-test it in your plan
Set aside an evening with last year's deals and your billing data. The last column shows what earns a 3 or 4.
| Module question | What to check | Evidence for a 3 or 4 |
|---|---|---|
| How pricing and packaging were set and tested (m02-q9, all stages) | Ask the founders, sales and finance to write down the model and its reason, separately | The answers match, and the plan dates the last buyer test and what changed |
| A bill the buyer can predict (m02-q9) | Ask a recent buyer to estimate their first-year bill from your pricing page | Their estimate lands in your quoted range, and customers are warned before a bill jumps |
| The motion the model has to fund (m01-q5 at Seed and A, m01-q8 at B) | First-year contract value against what it costs to win a customer in the main motion | Last year's deals show the motion pays back at this price point, with each motion shown separately at Series B |
| Channel and model, model and market (NEW-01c, Series B) | The four-fits scorecard from module 01 | Both pairs that involve the model are rated, with evidence and a market signal to watch |
| From free to paid (m03-q13 at Seed, m03-q14 at A and B) | Where free or trial users first meet the paid model | A defined upgrade trigger, conversion by cohort, and rules for which accounts go to sales |
| Cost to serve AI (m02-q9) | What each AI feature costs you per customer, against what that customer pays | Margin on AI features by customer, with a cap or alert on every metered line |
| Discount spread (NEW-02d, Series B) | Discounts on the same package, by region and team | A quarterly report that says whether the widest spreads trace to the model. If buyers want discounts because they can't forecast the bill, fix the model first. Guide E covers the rules |
What differs by stage
| Stage | What we'd expect the plan to say |
|---|---|
| Seed, typically under $1M ARR | One model, explained in a sentence, that a buyer can say yes to without a call. The 2026 survey found companies under $5M ARR were the most likely to charge flat fees. If you sell AI features, know what they cost to serve from the first customer |
| Series A, typically $1M to $5M ARR | A way for revenue to grow inside each account, such as more seats, more usage or a higher plan. Check that the model still fits the motion the first sales hires run. A model only the founders can explain is hard for a new rep to sell |
| Series B, typically $5M to $20M ARR | An owner for pricing, named in the plan, with the model written and agreed (a 3 on module 02). The 2025 survey describes $5M to $20M ARR as a no-man's land where nobody owns pricing. Where the company runs more than one motion or product, each may need its own model |
Sources
- Kyle Poyar, The state of B2B monetization in 2026, Growth Unhinged, May 2026
- Kyle Poyar, 2025 State of B2B Monetization, Growth Unhinged, June 2025
- Growth Unhinged, free-to-paid conversion report, February 2026
- Growth Unhinged, a guide to reverse trials, April 2026
- Bessemer Venture Partners, The AI pricing and monetization playbook, February 2026
- Georg Tacke, Simon-Kucher, 5 pricing models companies need to consider when innovating, July 2018
- Madhavan Ramanujam and Georg Tacke, Monetizing Innovation, Wiley, 2016
- Lenny's Podcast, Madhavan Ramanujam on pricing and scaling your AI product, July 2025
- Transcript of the same episode, Rosetta
- Brian Balfour, Why product market fit isn't enough (Four Fits), 2017
- Company pricing pages and announcements as linked above, read on 4 November 2026
Want a second view
on next year's plan?
More red than you expected? We can help fix it, working inside the business with your team. We'll reply within one working day.
Let's talkInvesting in a company that needs this? See how we work with VC and PE funds.