.png)
Frameworks, core principles and top case studies for SaaS pricing, learnt and refined over 28+ years of SaaS-monetization experience.
Thank you! Your submission has been received!
Oops! Something went wrong while submitting the form.
How to Design a Viral Freemium Strategy Lessons from Notion's 10B Growth Engine
On 19 May 2020, Notion made a decision that would have looked uncomfortable on a conventional SaaS pricing spreadsheet: it stopped charging individual users $5 a month for unlimited content. A solo user could now keep creating in Notion indefinitely without paying. Four years later, Notion said it had passed 100 million users. In July 2022, an employee share tender took place at the same $10 billion valuation as its previous financing round; by January 2026, a new $270 million tender valued the company at $11 billion.
Valuation does not prove that freemium caused the growth. Nor can outsiders reconstruct Notion's free-to-paid conversion rate: Notion remains private, does not file a public 10-K or conduct public quarterly earnings calls, and its October 2021 SEC Form D declined to disclose revenue. What the public record does show is a deliberate sequence. Notion first widened the amount of useful work a person could do for free, then moved the reasons to pay towards persistent teamwork, administration, security and, increasingly, AI.
Monetizely's position is that Notion's freemium engine worked because Free became a complete individual product rather than a crippled trial. Payment was pushed downstream to the point where individual utility becomes organisational coordination. In 2026, Notion should make that boundary even cleaner: remove the opaque free-team block limit, keep persistent member seats as the primary paid meter, and use credits only for autonomous work whose consumption can grow without another human seat.
Notion made the free user part of its distribution system
The important lesson from Notion is not "give more away". Plenty of products do that and build large populations of people who never pay. A successful freemium system has to connect product use, distribution and conversion in the right order.
Monetizely's 5-Step Pricing Framework provides that sequence. As developed in Monetizing Agentic AI, it begins with Goals & Segmentation, which identifies whom the product serves and what commercial result pricing must support. Packaging decides what each segment receives. Pricing Metric chooses the unit that makes the customer's bill grow. Rate Setting establishes the actual price only after the package and meter make sense. Operationalization then turns the design into working entitlements, upgrade rules, billing, limits and usage controls. Freemium puts unusual pressure on all five steps because a free package is simultaneously a product experience, an acquisition channel and the starting point of a paid account.
Applied to a viral product, the sequence is concrete:
- Goals & Segmentation: distinguish the solo user discovering the product, the small team beginning to collaborate and the organisation that needs governance.
- Packaging: give the free user enough functionality to finish a real job, while reserving capabilities created by sustained organisational use for paid plans.
- Pricing Metric: make the bill rise with something customers recognise as growth in value, such as persistent members, rather than an arbitrary count of basic actions.
- Rate Setting: choose the £ or $ amount only after the conversion boundary is clear.
- Operationalization: ensure invitations, guests, upgrades, proration and AI consumption behave exactly as the pricing promise suggests.
Notion's May 2020 reset is instructive because it changed Packaging before Rate Setting. Before the change, personal users paid $5 a month for unlimited content. Afterwards, individuals could create unlimited content for free indefinitely, while the paid Personal Pro plan remained $5 monthly or $4 a month on annual billing and added benefits such as larger file uploads, unlimited guests and longer version history.
The free user could therefore build something worth keeping before encountering a bill. More importantly, the artefact created inside the free product could leave the user's private workspace. Notion supported public sharing and editing, and by December 2022 its expanded Free plan allowed up to 10 guests while team workspaces could try collaboration without a time limit or credit card.
Akshay Kothari, Notion's co-founder, made the intended loop unusually explicit in December 2022: "Our hope is that people who discover Notion's value in its free offering will also introduce our product to their colleagues at work."
Templates amplified the mechanism. Notion reported in June 2023 that almost 11 million people had made 51 million template duplications during the preceding year; its rebuilt gallery expanded from roughly 600 templates to more than 5,000, supplied by more than 2,000 creator profiles. Its API had also attracted more than 30,000 people to the developer community by its March 2022 general-availability release. A user was no longer merely consuming software. Creating a public template, sharing a project page or connecting an application could expose another person to Notion.
Peer-reviewed research supports the underlying economics without implying that every product should become free. Niculescu and Wu's 2014 Information Systems Research model found that stronger word-of-mouth effects can make free seeding more attractive, while the optimal policy still depends on how the free version affects learning and later demand. Research in Management Science on referral programmes similarly finds that incentives can increase referrals but can also damage engagement when the referral requirement becomes too intrusive.
Notion is not alone in using a free product to let users carry distribution. The comparison matters because it separates a viral mechanism from mere free access.
| B2B SaaS product | What the free user spreads | How paid expansion occurs | Primary-source signal |
|---|---|---|---|
| Notion | Pages, templates, shared workspaces and invitations | Persistent team membership, administration, security and AI | Notion reported 100M users in August 2024 after reaching 1M in 2020. |
| Dropbox | Shared files and folders | Greater capacity and paid collaboration | Dropbox's 2025 10-K says word of mouth, referrals and sharing drive sign-ups; more than 90% of revenue comes through self-service. |
| Asana | Tasks, projects and invitations | Team adoption and larger organisational deployments | Asana's FY2026 10-K says its product-led motion relies in part on free users convincing other users to adopt; it reported more than 180,000 paying customers at 31 January 2026. |
| Atlassian | Team use of free and entry products | Larger deployments and higher cloud ARR | Atlassian reported more than 300,000 customers at 30 June 2025, with free and starter users beyond the reported customer count; customers above $10,000 of cloud ARR represented the majority of cloud revenue. |
The pattern is not "free creates growth". Free becomes commercially powerful when use naturally introduces other people and those relationships later create reasons to buy. Dropbox offers the cautionary counterpoint: a September 2023 filing disclosed more than 700 million registered users but 18.17 million paying users, while noting that registered-user counts include duplicates. That roughly 2.6% account ratio is not a true conversion rate, but it shows why sheer free reach cannot be the business case.
Asana CEO Dan Rogers described the same product-led challenge in March 2026 as an effort to "redefine how users discover Asana and how quickly they realize value once they enter." Notion's stronger insight was to make discovery happen inside a usable product, then let completed work become the invitation.
A freemium strategy becomes easier to understand when we trace what the company has progressively stopped charging for and what it has chosen to monetise instead. Notion's pricing history shows a clear movement away from charging for basic creation and towards charging for sustained teamwork and higher-value work. Its one significant detour was the first version of Notion AI pricing.
The direction matters more than any individual price point. Notion's freemium engine grew stronger as the company removed basic personal creation from the paywall, while its paid proposition became more closely associated with sustained collaboration, controls and now machine-performed work.
The current seat logic is also relatively easy to explain. Members occupy paid seats on paid plans, while guests can collaborate on specific pages without becoming full members; additions during a billing period can create prorated charges. A customer therefore has a recognisable commercial event when another person becomes a persistent member of the workspace. One part remains awkward. A Free workspace with one person can create unlimited pages and blocks, but a Free workspace with multiple members is subject to a block limit. Notion's public pricing page describes the limitation without giving the buyer a simple member-based point at which Free ends.
Blocks are meaningful to Notion's product architecture. They are not a good customer-facing definition of organisational maturity. A three-person start-up knows it has three team members; it should not have to forecast how many paragraphs, database items and page elements those three people will create before a paywall appears.
Notion gets packaging mostly right but its first AI meter exposed a weakness
The 5-Step Pricing Framework makes Notion's strengths and weaknesses clearer when we isolate the three decisions most relevant to this teardown: Packaging, Pricing Metric and Operationalization. Our grading is intentionally demanding because a viral model has to minimise friction at the same time as it protects revenue.
The scorecard says something more useful than "Notion has good pricing". Notion's packaging architecture has been ahead of its choice of meter at several moments. The company understood early that individual creation should be easy to enter and that organisations would pay for deeper needs; it took longer to find the right way to charge when AI began doing work independently of human seats.
The February 2023 AI launch illustrates the problem. Notion offered 20 free responses per member and then priced AI at $10 per member per month, discounted 20% under annual billing. For a 100-person workspace, the meter could therefore ask the customer to buy AI across human seats even when perhaps 20 people used it heavily and another 80 barely touched it.
Bundling AI into Business and Enterprise in May 2025 improved Packaging because customers no longer had to make a second purchase decision for every AI-enabled employee. Yet bundling cannot solve the next problem by itself: an autonomous agent can execute far more work without the company hiring another person. A pure seat meter eventually disconnects revenue from both Notion's cost and the customer's output.
Competitors face the same transition. Atlassian CEO Mike Cannon-Brookes said in an October 2025 earnings release that customers were choosing Atlassian as their "strategic AI platform" as the company expanded AI across its cloud products. Once AI becomes part of the operating system for work rather than an occasional writing feature, pricing has to distinguish human access from machine execution.
Autonomous work needs usage pricing without sacrificing the seat
Notion's Custom Agents make AI pricing directly relevant to the freemium design. Its February 2026 release describes agents that can operate without manual prompting, start from schedules or triggers, run continuously and work across Notion and connected services including Slack, Mail, Calendar and other tools connected through MCP. That behaviour is economically different from a human asking an AI assistant to summarise a paragraph.
The Agentic Monetization Spectrum (AMS) helps decide when that difference should alter the meter. AMS assesses an AI product on three dimensions. Zero-human ability asks how much useful work can finish without a person prompting every action. Operational domain asks how broadly the software can act across processes and systems. Output/cost ratio asks how much economically useful output the system can produce relative to its marginal execution cost. As scores rise, pure per-seat pricing becomes less defensible because work can scale even when human headcount does not. The meter should therefore move closer to autonomous output or consumption without discarding seats where seats still describe the human collaboration value.
Notion's Custom Agents score high enough to cross that line.
| AMS dimension | Score | Evidence and implication |
|---|---|---|
| Zero-human ability | 5/5 | Custom Agents can execute from triggers or schedules without repeated manual prompts and can operate around the clock. |
| Operational domain | 4/5 | Agents can act inside Notion and across connected work systems, but remain bounded by configured permissions and available integrations. |
| Output/cost ratio | 4/5 | Notion said in April 2026 that changes made Custom Agents 35%-50% cheaper to run, with newer models using up to 10x fewer credits for some work; credits are sold at $10 per 1,000. |
| Total | 13/15 | A human seat still fits collaboration, but autonomous execution warrants a second usage meter. |
A 13/15 AMS score argues against replacing Notion's seats with usage pricing. The persistent member remains the best primary meter because Notion's core product is still a shared workspace used by people. The score instead supports a specific architecture: seats for human membership, included credits inside the appropriate paid package, and credit consumption for autonomous execution above that allowance.
Notion has already built much of the operating machinery required to make that credible. By May 2026 it said more than one million Custom Agents had been created in the first two months, while administrators could monitor credits and set agent limits; agents pause when available credits are exhausted rather than silently generating an uncapped bill. Those controls matter because usage pricing without visibility turns an aligned meter into a procurement problem.
Workers push the same logic further. Notion's current documentation says their beta remains free through 10 August 2026 and starts consuming Notion credits on 11 August. A published example estimates typical cost at about $0.0023 per run; 1,000 credits, priced at $10, equate to roughly 4,348 typical runs, although actual consumption varies with the work performed.
Brian Emerick, Technical Program Manager at Vercel, described the appeal of Custom Agents as "a great way to spin up something quickly and elegantly without over-cooking a custom solution." Ease of creation is precisely why the usage layer has to be separated from the human seat. Ten employees can create hundreds of recurring machine workflows; charging only for the ten employees would eventually underprice the activity those agents perform.
Notion's next reset should make the first paid moment unmistakable
Notion should now complete the move it began in 2020. The next pricing reset should eliminate the Free plan's block-based team limit and put the first paid boundary on persistent internal membership.
Monetizely's recommended design is one clear architecture, not a menu of alternatives. Free should allow up to three persistent members to use the core workspace without a block limit, alongside generous guest sharing. The fourth persistent member should trigger Plus. Plus and Business should continue to use the paid member seat as their primary metric, preserving the commercial logic customers already understand.
Business should then include a meaningful workspace pool of agent credits linked to its paid footprint. Autonomous Custom Agent and Worker execution above the included pool should consume additional credits at the published usage rate, while Enterprise should use the same seat-first structure with contracted credit pools, governance and administrative controls. The exact included pool needs internal cost and usage data that Notion does not publish, but the meter itself should be unambiguous.
Three free persistent members would do something the block limit cannot: let a user experience the collaborative product before paying while defining conversion in a unit every buyer understands. A founder inviting two colleagues knows exactly what changed when a fourth person joins. No one needs to count blocks.
The broader design principle is visible throughout Notion's history. Free should remove friction from the behaviour that spreads the product. Paid packaging should capture the organisational value created after that spread. Usage pricing belongs only where machine activity can scale separately from the humans using the workspace.
For operators building their own freemium engine, five decisions follow from that position:
Measure free acquisition by the paid accounts it creates downstream, not by free registrations. Track which paying organisations began with a shared free artefact, invitation, template or other user-generated entry point. A million isolated free accounts are less valuable than 100,000 accounts that regularly introduce new organisations to the product.
Make the conversion event correspond to a change the customer can see in the business. A fourth persistent collaborator, a regulated workflow or an autonomous process is legible. A hidden object count is not. Customers should be able to predict tomorrow's bill from tomorrow's operating plan.
Build separate unit economics for human activity and autonomous activity before expanding AI. Seat economics should answer what another employee costs and contributes; agent economics should answer what another thousand runs, resolutions or completed tasks cost and contribute. Mixing both inside one seat hides whether AI expansion improves or destroys gross margin.
Treat free infrastructure spend as customer acquisition investment and demand evidence that it earns its keep. Dropbox's September 2023 disclosure of more than 700 million registered users against 18.17 million paying users shows why scale alone is an inadequate scorecard. Referral-originated cohorts should eventually outperform their infrastructure cost through conversion, expansion or materially cheaper acquisition.
Protect the product behaviour that creates distribution even when short-term monetisation pressure rises. Notion's decisive 2020 move was to remove the individual content ceiling, not tighten it. By August 2024, the company said it had passed 100 million users, and its January 2026 tender placed the company at an $11 billion valuation. Those numbers do not establish causal ROI, but they are consistent with the long-term wager that a useful free product can become a distribution channel before it becomes an invoice.
Monetizely's view of what right looks like in 2026 is therefore clear: the full individual job should be free, persistent organisational collaboration should be paid primarily by member seat, and autonomous machine work should consume usage only after an included allowance. Notion is already close to that design. Removing the block cap would finish the job.
Assumptions
AMS scores use a 1-to-5 Monetizely judgement based on publicly documented product behaviour; they are not Notion-reported scores. The proposed three-member Free boundary and included Business credit pool are Monetizely recommendations, not announced Notion terms. $10 billion and $11 billion figures refer to private-company tender valuations, not revenue. Notion is private and has no public 10-K or quarterly earnings-call transcript; historical Notion prices therefore use its dated first-party pricing and release records alongside its SEC Form D. Prices are in US dollars unless GBP is stated, and current Plus and Business prices refer to the annual-billing view.
Footnotes
https://www.amazon.com/Monetizing-Agentic-AI-Handbook-Transformation/dp/B0H7Z13VKJ/
Notion, pricing release, 19 May 2020: https://www.notion.com/releases/2020-05-19
Notion, "100 Million of You," 3 September 2024: https://www.notion.com/es-es/blog/100-million-of-you
Notion, January 2026 employee tender announcement: https://www.notion.com/blog/gic-sequoia-index-purchase-notion-shares
U.S. Securities and Exchange Commission, Notion Labs, Inc., Form D, filed 13 October 2021: https://www.sec.gov/Archives/edgar/data/1781814/000178181421000002/xslFormDX01/primary_doc.xml
Notion, expanded Free Plan and Business Plan announcement, 7 December 2022: https://www.notion.com/blog/notion-introduces-business-plan-and-expanded-free-plan
Notion, pricing-plan release, 7 December 2022: https://www.notion.com/releases/2022-12-07
Notion, Template Gallery announcement, 21 June 2023: https://www.notion.com/blog/new-notion-template-gallery
Notion, API general availability announcement, 2 March 2022: https://www.notion.com/blog/api-ga
Marius F. Niculescu and D. J. Wu, "Economics of Free Under Perpetual Licensing," Information Systems Research, 2014: https://pubsonline.informs.org/doi/abs/10.1287/isre.2013.0508
Management Science, "Social Referral Programs for Freemium Platforms": https://pubsonline.informs.org/doi/10.1287/mnsc.2022.4301
Dropbox, Inc., Form 10-K for fiscal 2025: https://www.sec.gov/Archives/edgar/data/1467623/000146762326000008/dbx-20251231.htm
Asana, Inc., Form 10-K for fiscal year ended 31 January 2026: https://www.sec.gov/Archives/edgar/data/1477720/000147772026000021/asan-20260131.htm
Atlassian Corporation, Form 10-K for fiscal year ended 30 June 2025: https://www.sec.gov/Archives/edgar/data/1650372/000165037225000036/team-20250630.htm
Dropbox, Inc., SEC filing for quarter ended 30 September 2023: https://www.sec.gov/Archives/edgar/data/1467623/000146762323000051/dbx-20230930.htm
Asana, fiscal 2026 fourth-quarter earnings-call transcript, 2 March 2026: https://investors.asana.com/static-files/24e55460-efa4-47d0-a778-29d3b5366936
Notion AI pricing release, 22 February 2023: https://www.notion.com/en-gb/releases/2023-02-22
Notion Plus pricing update, 26 June 2024: https://www.notion.com/en-gb/releases/2024-06-26
Notion AI Business and Enterprise packaging update, 13 May 2025: https://www.notion.com/en-gb/releases/2025-05-13
Notion Custom Agents release, 24 February 2026: https://www.notion.com/releases/2026-02-24
Notion Custom Agent pricing and efficiency update, 14 April 2026: https://www.notion.com/en-gb/releases/2026-04-14
Notion, current UK pricing page, accessed 10 August 2026: https://www.notion.com/en-GB/pricing
Atlassian, first-quarter FY2026 earnings release filed with the SEC, 30 October 2025: https://www.sec.gov/Archives/edgar/data/1650372/000165037225000064/ex991q1fy26.htm
Notion, current Custom Agents pricing, accessed 10 August 2026: https://www.notion.com/product/custom-agents
Notion Custom Agents usage-control release, 5 May 2026: https://www.notion.com/releases/2026-05-05
Notion, "Understand pricing for Workers," accessed 10 August 2026: https://www.notion.com/en-gb/help/understand-pricing-for-workers