Self-Serve SaaS Conversion
Compare freemium, trials, card requirements, and activation using vendor surveys, transaction logs, and randomized experiments to establish SaaS forecasting baselines.
Written and reviewed by Eric Han · Co-founder & CTO, reopt
Key takeaways
- Overall median free-to-paid conversion is 8% (ChartMogul 2026). Good ranges are 3–5% for freemium and 4–6% for no-card trials.
- Card-required products show higher trial-to-paid but lower visit-to-trial conversion. Multiplying stage rates gives 10.5 paying customers per 1,000 visits for card-required trials versus 3.6 for all trials in ChartMogul. Grade D data with unclear methods give 12.2 for card-required trials versus 15.5 for trials without a card. We cannot establish which produces more customers overall.
- A randomized trial-length study found seven days better than 30 (15.36% versus 14.67%, one company); another found seven better than three. Optimal length depends on the range tested.
- Activation averages 34%, with a 25% median, but companies define activation differently.
- Observational research finds shorter customer lifetimes among free-trial acquisitions. Do not choose a model on conversion alone.
Free-to-Paid by Model
Both surveys use self-selected samples and six-month windows. Lenny's denominator is accounts; ChartMogul's is leads/free signups (grade C). ChartMogul defines “good” as the 50th percentile and “great” as the 75th. Lenny's thresholds were not confirmed.
| Model | ChartMogul 2026: good / great | Lenny 2023: good / great |
|---|---|---|
| Overall | Median 8% | — |
| Freemium | 3–5%, 8–12% | 3–5%, 6–8% (self-serve) |
| Freemium with sales assistance | — | 5–7%, 10–15% |
| Trial | 4–6%, 10–15% (no card) | 8–12%, 15–25% |
| Trial, card required | 25–35%, 50–60% | — |
- ChartMogul / Growth Unhinged / ProductLed: 200 software products, January 2026.
- Lenny's Newsletter: more than 1,000 products, including OpenView and Pendo sources; predominantly B2B, 2023.
- ProductLed: free-to-paid about 9%; median 10% at $1K–$5K ACV; PQL-based trial conversion 25% (more than 600 SaaS companies; year unclear).
The distribution is wide. In ChartMogul, 20% of trial products convert below 2.5%, while 23% exceed 25%. Because respondents are mature enough to measure conversion, early startups should plan around outcomes near the lower end of the interval (author judgment; no evidence for a numerical discount factor).
Card Requirements: No Single Answer
All these data are observational. Companies select their own card policy; we found no randomized test of the requirement itself.
- ChartMogul: card required10.5customers
- ChartMogul: all trials3.6customers
- First Page Sage organic: card12.2customers
- First Page Sage organic: no card15.5customers
- First Page Sage paid: card11.2customers
- First Page Sage paid: no card12.4customers
| Metric | Card required | No card | Source |
|---|---|---|---|
| Trial → paid | 39.0% | 12.0% | Recurly 2022, B |
| Trial → paid | About 30% | 4–6% | ChartMogul 2026, C |
| Trial → paid | 48.8% | 18.2% | First Page Sage, organic traffic, D |
| Visit → trial | 3.5% | 4.5% (all trials) | ChartMogul, C |
| Visit → trial | 2.5% | 8.5% | First Page Sage organic, D |
| Paid per 1,000 visits | 10.5 | 3.6 (all trials) | ChartMogul, calculated by multiplying stage rates |
| Paid per 1,000 visits | 12.2 | 15.5 | First Page Sage organic, calculated |
| Paid per 1,000 visits | 11.2 | 12.4 | First Page Sage paid, calculated |
Card-required products have higher observed trial-to-paid and lower visit-to-trial rates, but the multiplied outcomes conflict. Simulate both paths separately and choose using your own experiment.
ChartMogul's 4.5% and 3.6 customers cover all trials, including the 20% requiring a card. No-card-only visit-to-signup is undisclosed, so the end-to-end rate for trials without a card cannot be calculated. Recurly includes B2C, and 87.4% require a payment method. The nearest randomized evidence is automatic renewal versus automatic cancellation for European newspaper subscriptions: automatic renewal deterred 24–36% of potential subscribers but increased persistence among signups sevenfold. This is neither a test of card requirements nor B2B SaaS.
Trial Length: Experiments and Observations Conflict
| Evidence | Shorter | Longer | Nature |
|---|---|---|---|
| Yoganarasimhan et al., 2023 | 7 days 15.36%; 14 days 14.96% | 30 days 14.67% | 337,724 randomized users, A |
| Recurly 2022 | 7 days or fewer: 40.4% | More than 61 days: 30.6% | Observational transaction logs, B |
| RevenueCat 2026 | 4 days or fewer: 25.5% | 17–32 days: 42.5% | Observational app transaction logs, B |
In the randomized study, giving everyone seven days rather than 30 increased subscriptions by 5.59% in test data (4.34% in training data). The raw difference is 0.69 percentage points. Personalization increased subscriptions more than 6.8% versus a uniform 30-day trial, and subscription-optimized personalization raised long-term revenue 11.61%. This comes from one anonymous large SaaS provider in 2015–16.
A Frontiers in Psychology 2025 randomized study of 680,588 image-editing SaaS users found seven days versus three increased trial starts 11.1% and delayed conversion about 42%, with no significant immediate-conversion difference (a correction exists; unequal allocation). Shorter is therefore not universally better. RevenueCat trial lengths are chosen by apps and confounded with category. This book prioritizes experiments while recognizing that their findings may not generalize to other settings.
Activation
| Source | Mean | Median | Grade |
|---|---|---|---|
| Lenny 2022, overall | 34% | 25% | C |
| Lenny 2022, SaaS | 36% | 30% | C |
Activation events differ across products. Userpilot industry figures are excluded because the original could not be verified and sample descriptions conflict. ProductLed's 34% of PLG companies tracking activation is a measurement-adoption rate, not an activation rate.
Include Retention
An observational study found free-trial customers stayed about one-third as long as regular customers and generated about 59% less total profit (Datta et al., JMR 2015, European digital TV, via press coverage). It is neither SaaS nor randomized, but it cautions against choosing a model based only on conversion counts.
Simulation Inputs
| Stage | Input | Evidence and conditions |
|---|---|---|
| Visit → trial | Card required 3.5%; all trials 4.5% | ChartMogul 2026, C |
| Trial → paid, no card | Good 4–6%, great 10–15%; mean 8.7%, | Fitting the 50th/75th percentiles is this book's calculation |
| Trial → paid, card required | 25–35%, center 30% | Same source, C |
| Freemium free → paid | 3–5% | Same source, C |
Do not use First Page Sage figures as inputs (D, directional comparisons only) or Userpilot industry figures.
Missing Evidence
- Randomized card-requirement tests and SaaS freemium field experiments.
- Recent ACV-band tables and independent collaboration, HR, and fintech data.
- Public Korean B2B SaaS trial/freemium conversion figures. Searches for flex, Channel Talk, Wrtn, Speak, and JANDI found none. Nearby evidence is collected in Korean Cases.
Effect Sizes of Ecommerce Levers
Distinguish evidence on speed, reviews, free shipping, discounts, and checkout UX, and examine realistic A/B test effects.
Early-Stage SaaS Conversion
Review retention, growth, and first-revenue evidence from prelaunch to $1 million ARR, and identify what evidence can and cannot justify discounting mature-company benchmarks.