Software Growth

Free trial

A free trial gives prospects full use of a paid product for a set time, usually 7 to 30 days, before they pay or lose access. Measure it with the trial-to-paid rate.

A free trial lets a prospect use your paid product for a fixed period, usually 7 to 30 days, before they have to subscribe or lose access. You measure it with one number: trial-to-paid rate = trials that became paying customers ÷ trials started.

It is the most common way into self-serve SaaS. In Kyle Poyar's 2026 survey of 200 B2B software products, run with ChartMogul and ProductLed, 57% used a free trial as the main entry point for new customers, more than twice the share using freemium (26%). The trial lets the product do the selling, and the deadline gives people a reason to decide.

How a free trial works

A trial has four moving parts, and each one is a design decision:

  • The entry. Do you ask for a credit card at signup or not? This decides who starts a trial and what happens at the end.
  • The clock. How many days, and when they start counting. Most products start at signup; some start on the first meaningful action.
  • The path to value. What the user has to do before the product proves itself: import data, invite a teammate, send a first invoice. This is onboarding, and the moment it works is the aha moment.
  • The end. When the clock runs out, the user pays, loses access behind a paywall, gets billed automatically, or drops to a free plan. Which one depends on the model you chose.

Almost everything that goes wrong with trials comes from a mismatch between the clock and the path to value.

Free trial vs freemium vs reverse trial

Four ways to let people try before they pay, shown on a timeline from day 0 to day 14 and beyond. A free trial without a card unlocks all paid features for 14 days, then the user pays or loses access. A card-required trial unlocks the same features, then the user is billed automatically unless they cancel. A reverse trial unlocks paid features for 14 days, then drops the user to a limited free plan. Freemium gives a limited free plan from day one, and users upgrade when they hit a limit.
Software Growth, based on definitions by Kyle Poyar (2026) and Elena Verna (2024). Schematic.

Three models compete for the same job of letting people try before they pay:

  • Free trial: full access for a limited time, then pay or leave. The deadline pushes a decision.
  • Freemium: a free plan with limits that never expires. Users upgrade when they hit a limit or need a paid feature.
  • Reverse trial: every new signup starts on a trial of paid features and drops to the free plan if they don't pay. Elena Verna, who leads growth at Lovable, calls it the best of both, and shows Canva, Asana, Notion and Airtable using it. Airtable's reverse trial runs 14 days and Asana's 30.

In Poyar's 2026 data, only 7% of products ran a reverse trial. He also found that free trials convert a little better than freemium, but that the gap disappears once you account for how many more people sign up for a free plan. So the model alone won't settle it.

What goes into the model matters more than the model itself.

Wes Bush / Freemium vs. Free Trial: How to choose the right model (May 30, 2023)

For bootstrapped founders, Rob Walling's default is a trial. In episode 754 of Startups for the Rest of Us (from 3:12) he argues that freemium is much harder to make work than founders expect, so if you are in doubt, skip it and run a free trial, where people have a time constraint to convert. His point is that you need to find out whether you built something people want and will pay for.

Opt-in vs opt-out: should you ask for a card?

The biggest design choice is whether you ask for a credit card up front.

  • Opt-in trial: no card required. Signup is easier, so you get more trials, but a smaller share convert because the user has made no commitment.
  • Opt-out trial: card required, and the user is charged automatically when the trial ends unless they cancel. You get fewer signups but a much higher conversion rate among those who start. Expect more refund requests, chargebacks and angry emails from people who forgot to cancel.

The data shows how big the gap is. Among the free trial products in Poyar's 2026 report, 20% required a card, and those converted at 30%, more than five times the rate of trials without one. That headline is easy to misread. A card form also turns away many of the people who would have started, so the conversion rate rises partly because the denominator shrinks.

You don’t need to ask for credit card upfront.

Rob Walling / Startups for the Rest of Us, episode 754 (February 4, 2025), from 3:12

Neither is right everywhere. Compare paying customers per 1,000 visitors, not the conversion rate alone, because the two models change the denominator. The worked example below shows why. Some products avoid choosing: one company in Poyar's follow-up report (March 18, 2026) put two buttons on its homepage, one for freemium and one for a 14-day card-required premium trial, and saw a 26% improvement in premium trials created.

If you do require a card, Paddle's guide to free trials gives the rule that keeps it fair: be upfront about when you will bill, and make cancelling easy. We'd add one more step: email a reminder a few days before the first charge. It may cost a few conversions, and it saves chargebacks and complaints.

How long should a trial be?

Two fictional products on a 30-day timeline. An invoicing tool reaches its aha moment, the first invoice paid, on day 2, so a 7-day trial fits. A team reporting tool needs data connected on day 3, the team invited on day 9 and the first report with real trends on day 16, so a 7-day trial ends before any value and a 21-day trial fits.
Fictional products. Schematic.

Match it to time to value. If a user can reach the point where the product is useful in one session, 7 days is plenty. If they have to import data, invite teammates or wait for results to accumulate, you need 14 to 30 days or the trial ends before they see anything.

Most companies land on two weeks. In Poyar's 2026 data, 62% of trial products used 14 days, 14% used 7 days and 14% used 30 days. Paddle's advice is to keep the trial as short as possible while still giving people time to learn the key features. Too long a trial delays payment and lets some people finish the job they came for before they ever pay.

A longer trial rarely fixes a weak first session. Fix onboarding first.

How to measure it

If 400 people start a trial in a month and 60 subscribe, your trial-to-paid rate is 15%. At a $39 plan that is $2,340 of new MRR. Track this as a trial conversion rate, and split it by acquisition channel, since visitors from a Google search behave differently from people sent by a friend.

Three habits keep the number honest:

  • Measure by cohort. Count conversions for the people who started a trial in a given month, not all conversions that happened that month. Lenny Rachitsky and Kyle Poyar's benchmark uses each signup cohort's first six months. See cohort analysis.
  • Track the step before it. Your activation rate, the share of trial users who reach the aha moment, tells you why the trial-to-paid rate moved.
  • Look one step further. For card-required trials, count customers still paying after the first billing cycle, since refunds and early cancellations eat into the headline rate.

What good looks like

Benchmarks are survey-based and self-reported, so treat them as a range to sanity-check against, not a target. Price and audience move them a lot.

  • Lenny Rachitsky and Kyle Poyar (2023), 1,000+ products: free trial products converted at 8 to 12% for "good" and 15 to 25% for "great" (source). Trials also drew fewer signups: a 5% sign-up rate, against 9% for freemium products.
  • Kyle Poyar (2026), 200 products: the median free-to-paid rate across all products was 8%, but the spread is wide. Among trial products, 20% converted below 2.5% and about 23% converted above 25%. Card-required trials converted at 30%.
  • Who you sell to: the 2023 data found lower conversion for products selling to larger companies and to developers. The median for developer-focused companies was 5%, half that of the rest.

Worked example: card or no card

Fictional example of 1,000 visitors to a $49 per month SaaS. With no card required, 8% start a trial (80 trials), 15% of those pay (12 customers) and 11 are still paying after month 1, giving $539 of MRR. With a card required, 3% start a trial (30 trials), 30% pay (9 customers) and 7 are still paying after month 1, giving $343 of MRR.
Fictional example. Numbers are illustrative, not benchmarks.

Take a fictional scheduling tool at $49 a month, and send the same 1,000 visitors to two versions of its trial.

  • No card required: 8% start a trial, so 80 trials. 15% convert: 80 × 0.15 = 12 customers. One cancels in the first month, leaving 11. New MRR: 11 × $49 = $539.
  • Card required: 3% start a trial, so 30 trials. 30% convert: 30 × 0.30 = 9 customers. Two ask for a refund or cancel after the first charge, leaving 7. New MRR: 7 × $49 = $343.

The card version doubles the trial-to-paid rate (30% vs 15%) and still brings in less revenue, because it lost more visitors at the door than it gained at the end. The deciding number is how many visitors the card form turns away. If 5% of visitors still started a card-required trial, that would be 50 trials and 15 customers. With the same refund share (2 in 9), about 11.7 stay on average (15 × 7/9), or roughly $572 of MRR. The card version would then pull slightly ahead.

Now add an AI product's running costs. Say each trial costs $10 in model usage. The no-card version spends 80 × $10 = $800 to win $539 of MRR, so it takes about 1.5 months of revenue to pay back. The card version spends 30 × $10 = $300 for $343 of MRR, under a month. With expensive trials, the card requirement can win even when it brings in fewer customers. Run the same numbers with your own signup rates before you choose.

Getting trial users to value

The trial is the window in which the product has to sell itself. Walling's diagnosis in episode 754 for trials that don't convert starts with onboarding: is anyone actually getting onboarded, and what are the one, two or three steps someone needs to take to get value? If people aren't taking them, reach out by email, in-app message or, for larger accounts, by hand. If people don't get onboarded, he says, there will be no paid conversions.

Poyar's survey of what actually moved conversion (March 18, 2026) points to a handful of levers:

  • Attract people who intend to buy. Signups from organic search, referrals, social media and AI assistants converted best; paid ads converted worst.
  • Shorten the path to value. Shorter onboarding, onboarding by use case and AI-assisted setup reduce friction. Interactive demos and clearer calls to action raise motivation.
  • Change the plan design, not just the price. Adjusting limits, feature gates and trial mechanics did more than raising or lowering prices.
  • Add humans where the deal is big. 80% of trial products had human touchpoints when an enterprise user started a trial. See dual funnel and product qualified leads.
  • Give activation an owner. Product teams usually owned onboarding, but sales or growth usually owned the conversion number. Poyar suspects the gap means activation never gets enough attention.

Free trials in the AI era (September–October 2026)

For AI products, a trial is no longer almost free to run. Every prompt a trial user sends costs real money. Poyar's 2026 report notes that supporting free users has become much more expensive because AI token costs have stayed high, and that AI-native products use free trials less often (43%) than classic SaaS (61%).

Elena Verna argues against reacting by cutting back what people can try. In a September 8, 2026 post, she writes that giving the product away is one of Lovable's biggest growth levers, even though every use costs money, and warns that the AI-era panic over gross margin has pushed companies to hide the very experience customers need before they will pay. In a SaaStr clip published September 29, 2026, she describes Lovable treating its freemium costs as marketing budget. For a trial, that means comparing the cost of trial usage with your customer acquisition cost, as in the worked example above.

Abuse is the second new cost. Bots, stolen accounts and serial trial-takers used to be a data problem.

In traditional SaaS, a fake signup mostly messed up your conversion metrics.

Elena Verna / What on earth is going on with AI pricing and packaging? (October 16, 2026)

In AI, she adds, it ruins your bottom line. The same post argues that AI companies have done a bad job explaining what their credits are worth. If you sell credits or tokens, the trial is the place to fix that: show what a typical allowance actually gets done.

The trial also doesn't end at the first payment. Poyar's analysis of ChartMogul data from 4,029 software companies (September 30, 2026) found that new customers were up to 10 times more open to upgrades in their first 30 days than in months four to eleven. One of his suggestions is a second trial: soon after someone buys, offer a 30-day trial of the next tier up. His summary:

the first 30 days are worth more than the next 300.

Kyle Poyar / The 30-day upsell window (September 30, 2026)

Practical steps if your trial triggers paid model calls (our suggestions, not the sources'):

  • Cap trial usage in units the user understands, such as projects, documents or minutes, rather than open-ended credits.
  • Require email verification, and consider a card for the most expensive features, so the cost of abuse stays bounded.
  • Track trial cost per paying customer next to your acquisition cost, and review it monthly.
  • Put the aha moment inside the cap. If users need more than the cap to see the product work, the cap is too tight.

Common mistakes

  • Starting the clock before the user has done anything. Some products start the trial on first meaningful action instead of signup.
  • Offering no way to extend. A "need more time" button that grants 7 more days to engaged users is cheap to offer.
  • Judging opt-in vs opt-out on conversion rate alone instead of paying customers per visitor.
  • Hiding when the card will be charged. It produces chargebacks and refund requests, and it damages trust you need for renewals.
  • Gating the aha moment behind the paid plan to save on AI costs, so trial users never see what they would pay for.
  • Dead-ending at expiry. If you want a softer landing, look at a reverse trial or a limited free plan (freemium). People who don't convert can also come back later through a win-back campaign.

Sources

Back to the SaaS glossary