Freemium vs free trial: choose by paying customers, not sign-ups
Freemium raises sign-ups. A trial raises free-to-paid. Compare both by paying customers per 1,000 visitors, then test the switch by cohort.
Pick freemium or a free trial by one number: paying customers per 1,000 visitors in the same cohort. Picture a founder who went freemium last quarter: sign-ups doubled on the dashboard, and new paying customers stayed flat. Both rates did what the models predict. The free plan pulled more people in, and a smaller share of them paid.
Sign-up rate and free-to-paid conversion move in opposite directions when you change the free model. Judge the two together, read them by cohort, and keep activation and refunds as guardrails.
Freemium vs free trial: the difference
The five common models differ in what a new user gets before paying and where the payment decision sits.
| Model | What a new user gets | When they pay | What it raises |
|---|---|---|---|
| Freemium | A permanent free plan with limits on usage, seats or features | When they hit a limit or need a paid feature | Sign-ups and top-of-funnel volume |
| Free trial, no card | Full access for a fixed period, no payment details | At the end of the trial, by entering a card | Urgency to decide, and a higher free-to-paid rate than freemium |
| Credit card required trial | Full access for a fixed period after entering a card | When the trial ends, unless they cancel | Free-to-paid, at the cost of fewer sign-ups |
| Reverse trial | Paid features for a fixed period, then a free plan | When they miss the paid features on the free plan | Awareness of paid features among free users |
| Ungated freemium | Use of the product before creating an account | After sign-up, when they hit a limit | Use before sign-up; in figure 1, fewer sign-ups than gated freemium but more payers |
Elena Verna's definition on the Amplitude blog is the clean one: "A reverse trial gives customers access to paid features and puts them on a freemium plan when the trial ends."
Ungated freemium is a sizable slice of the freemium group. In the ChartMogul Conversion Report, 38% of freemium products let users try the product before creating an account.
Write down which of the five a new user actually meets today, card step included. Describe it by the first session, not the pricing page.
What the 2026 benchmarks say
The most useful current source is the ChartMogul Conversion Report by Kyle Poyar, published in February 2026. The survey ran in January 2026 across 200 software products, collected through a Typeform survey, so the numbers are self-reported estimates. A typical respondent sits between $1 million and $10 million ARR, with average revenue of $50 to $249 per customer per month. Free-to-paid here means a lead or free sign-up who paid within 6 months.
Trials dominate the sample. In the survey, 57% of products use a free trial as the primary landing point, against 26% for freemium. Only 7% run a reverse trial. The most common trial length is 14 days, used by 62% of products. Among trial products, 20% require a card upfront.
The median free-to-paid rate across all products is 8%. That median mixes every model, card-required trials included, so it sits above the good ranges for freemium and no-card trials. Compare your rate with your own model's row:
| Model | Good free-to-paid | Great free-to-paid |
|---|---|---|
| Freemium, self-serve | 3% to 5% | 8% to 12% |
| Free trial | 4% to 6% | 10% to 15% |
| Reverse trial | 4% to 6% | 8% to 12% |
| Credit card required trial | 25% to 35% | 50% to 60% |
The report flags the reverse trial sample as small and the difference as not statistically significant.
The report's own verdict on the two models: "Free trials convert at slightly higher rates than freemium; however, this difference gets wiped out upon accounting for signup rates." Figure 1 shows the report's model per 1,000 website visitors, built from median survey rates. Freemium draws twice the sign-ups of a no-card trial and ends with more payers. The card-required trial draws the fewest sign-ups and ends with the most payers.
The 8% on the trial row is our own division of 3.6 by 45. Do not confuse it with the all-product median above. The computed rate also sits above the trial's good range, so use the visitor model to compare models, not as a target. In figure 1, the card trial wins because its sign-up loss is smaller than its conversion gain. Measure your own balance before you copy the result.
Elena Verna's 2022 numbers on the Amplitude blog run higher for trials. She writes that companies typically see 15% trial-start-to-paid on no-card trials. On freemium, she typically sees about 5% from a free plan to paid. On card-required trials, she writes that as many as 70% to 80% of starters convert. She also warns that "as many as 80% of people will drop off at the credit card screen."
Verna's figures are one practitioner's observations from 2022. ChartMogul's come from a 2026 survey with a fixed question. Different year, different method: do not plot them as one series. The sources agree on direction and disagree on level, and the ranges overlap across models. A benchmark tells you whether your number is sane. It cannot pick the model for you.
Why sign-ups and free-to-paid both mislead
Each rate has its own denominator. Sign-up rate divides sign-ups by visitors. Free-to-paid divides payers by sign-ups. Change the free model and both move at once, in opposite directions.
Go freemium and the sign-up rate climbs. The new sign-ups include people who never meant to pay, so free-to-paid drops. The sign-up dashboard celebrates and the conversion dashboard panics. Neither one is reporting the result.
Multiply the two rates and you get payers per visitor, the number the model actually changes. Read it as paying customers, and revenue, per 1,000 visitors. Group visitors into weekly cohorts by first visit. Read every cohort at the same fixed window after sign-up.
The card requirement shows the trap most clearly. Conversion jumps because only high-intent people start the trial. The ChartMogul report warns that a card requirement "could blunt signups, hurting the overall number of paying customers, unless the product warrants the friction of adding a credit card."
Track revenue per payer next to the count. A generous free plan pulls in solo users who later buy the cheapest tier. Payer count rises while revenue per payer falls, and the per-visitor revenue line catches it.
Keep activation rate as a guardrail on every read. When payers per visitor rises and activation falls, you are charging people before they see value. Expect churn in the second or third month.
How to choose: three questions
The ChartMogul report frames the choice around what you optimize for, when users reach value and what a free user costs. We turned that into three questions you answer from your own event data, not from a competitor's pricing page.
1. Can a new user reach value alone inside the trial length?
Pull the median time from sign-up to the first value event for users with no sales help. Compare it with your trial length. If most need longer, the clock runs out before the product makes its case. The time to value post covers how to define that event.
Elena Verna notes on the Amplitude blog that in B2B, time as a value metric favors smaller businesses: "What takes a startup a day to accomplish might take an enterprise a month." If you sell to both, split the time-to-value read by company size. A "no" here points to freemium or a reverse trial.
2. Does each free user cost real money?
List what a free user costs you: compute, storage, support tickets, fraud review. For a simple web tool that cost sits close to zero. For an AI product, every generation burns compute.
When free usage costs real money, a trial caps the exposure by date. If you keep a free plan, cap or price the costly part. On Pixel+, an AI photo and video editing app, Kreto redesigned onboarding and built a paywall architecture split between subscription and token-based models.
In any subscription and token hybrid, tokens price the expensive action and the subscription covers ongoing access. Inside a mobile app, the same choice shows up as a hard paywall vs soft paywall decision.
3. Do free users bring in new users?
Count the sign-ups that arrive through an existing free user: invites, shared files, public pages, exported work that carries your name. If that share is material, the free plan is an acquisition channel through a growth loop. A trial would cut that loop off.
Verna adds a catch: "Users on a free plan often lack awareness of your paid features." A loop full of free users who never see the paid tier grows usage, not revenue.
When a reverse trial fits
Test a reverse trial when question 1 rules out a plain trial and free users never find the paid features. Verna's guidance is to "analyze product usage to understand when users typically need more advanced features," then time the paid period around that point.
Whatever you pick, make the free part decisive. Wes Bush of ProductLed, quoted in the ChartMogul report, says: "The worst middle ground is a timid free experience that neither drives adoption nor creates urgency to buy."
How to test a switch
Run the switch as a cohort test, not a launch. Figure 3 shows the funnel you read and the one number you decide on.
- Baseline first. Read four weekly cohorts on your current model: visitors, sign-ups, activated users, payers and revenue.
- Fix the read window before you start. Pick 30 or 60 days after sign-up and never move it mid-test. A free plan has no deadline, so its payers arrive later than a trial's, and a 30-day window understates freemium. Decide at the window, then confirm freemium cohorts again at 90 days.
- Change one thing. Switch the model, the card requirement or the trial length, never two at once. A card requirement is its own test.
- Read payers and revenue per 1,000 visitors by segment. Split self-serve from team accounts, and split by channel. Paid social and organic search traffic react differently to a card screen.
- Watch the guardrails. Track activation rate, refunds and month 1 to 3 retention. For card trials, the ChartMogul report points to Canva, which "tells prospective trial-ers that they’ll get a reminder 7 days before their Canva Pro trial ends." Skipping that, the report says, risks "fewer trial signups, worse month 1-3 retention, and higher refund rates as people forget to cancel."
Kreto runs this kind of cohort test across sign-up, activation and paywall as part of product marketing. Whoever runs it, book the four baseline weeks before anyone touches the pricing page.
Common mistakes
- Copying a competitor's model. The ChartMogul report puts it plainly: "stop copying another company’s free model without understanding what it optimizes for." A board member's favorite example optimizes for its own loop and cost base.
- Judging by sign-ups. A free plan wins on sign-ups by design. The count says nothing about revenue.
- Judging by free-to-paid. A card-required trial wins on conversion by design. The rate says nothing about how many people you turned away.
- Getting the free plan size wrong. Make it so generous that nobody needs to pay, and revenue stalls. Make it so thin that nobody stays, and the loop dies. Decide which job the free plan does, then size it for that job.
- Reading freemium at day 14. Trial payers arrive at the deadline. Freemium payers arrive when they hit a limit, which takes longer. Read both models at the same window, and keep reading freemium cohorts after it closes.
FAQ
Is freemium better than a free trial?
Neither wins in general. Freemium brings more sign-ups, and a trial converts a larger share of them. Compare the two by paying customers and revenue per 1,000 visitors in your own cohorts. Freemium fits products where free users bring new users and cost little to serve. A trial fits products where a new user reaches value quickly and alone.
What is a reverse trial?
A reverse trial starts every new user on the paid tier for a fixed period. When it ends, the user moves to a free plan instead of losing access. The model suits products where free users stay but never discover the paid features. The trial shows them what they lose, and the free plan keeps them around.
Should a free trial require a credit card?
Test the card as its own change, separate from the model. A card raises the share of trial users who pay and cuts the number who start. Whether payers per visitor rise depends on how much high-intent traffic you have. If you add a card, send a reminder before the trial ends and watch refunds and early retention.
How long should a free trial be?
Fourteen days is the most common length, used by 62% of products in the ChartMogul report. Set yours by your time to value, not by habit. Measure how long a self-serve user takes to reach the first value event. Give most users enough time to get there, plus a few days of real use.
Pick the model by the customers who pay
You now have the three questions and a test plan. Run the baseline, change one thing, and read paying customers per 1,000 visitors at a fixed window, with activation and refunds as guardrails. If the switch touches pricing, onboarding and acquisition at once, talk to Kreto about running it as one funnel.