Trial to paid

What is a good trial to paid conversion rate?

Wide enough that the benchmark is only a sanity check, and you should get your own definition straight before you look at anyone else's number.

In a survey of more than 1,000 products run by Kyle Poyar, Lenny Rachitsky and the Pendo team, published in August 2023, a free trial converting 8 to 12 percent was called good and 15 to 25 percent great, while self-serve freemium was good at 3 to 5 percent. In the SaaS Conversion Report Kyle Poyar published with ChartMogul and ProductLed in March 2026, the median free-to-paid rate across 200 B2B products was 8 percent, and 14 days was the most common trial length.

Two companies can report 8 percent and 18 percent for the same funnel, depending on what they put in the denominator. So the first step is not a formality.

Step 1: Settle the denominator and the window

Write down three things and get finance to agree to them: what starts a trial, what counts as paid, and how long after the trial you keep counting.

A trial starts at the first session, not at the signup. If signups that never return sit inside your denominator, you are measuring your signup form and calling it your trial. Keep that gap as its own number, because it belongs to activation and has a different fix. Paid means the first successful charge, not the plan selection and not the card on file. The window is 30 days after the trial ends: any window that stays open improves the rate every time you look at it, which is why no two reports in the company agree.

What goes wrong here. Somebody reruns the report with a different window before a board meeting, the two numbers differ by 6 points, and the hour goes on arithmetic instead of the funnel. Publish the definition where the dashboard lives.

Step 2: Find the action that separates payers from leavers

Take everyone who started a trial in the last two completed quarters. Split them into the ones who paid and the ones who didn't. For each product event, compute the share of each group that did it at least once in the first seven days, and sort by the gap between the two shares.

You are not looking for the most common action. You want the action with the widest gap, done by most payers and few leavers. In most products two or three candidates come up, and one is obviously the thing the product is for: the first invoice sent, the first teammate invited, the first report that returned a real answer.

What goes wrong here. You find a correlation and ship it as a cause. Heavy users log in more, so "logged in 5 times" shows a huge gap and means nothing. The test is whether you can move the action for a person who wasn't going to do it, and have them still convert. Pick the candidate that is closest to the product's actual job, then prove it by making it easier for one cohort and watching the paid rate rather than the action rate.

Step 3: Count the steps between signup and that action, then cut them

Open a new account and do it yourself, with a stopwatch. Write every screen, field, confirmation email, permission request and empty state between the signup button and the action from step 2. Count them.

In almost every product I have done this in, the count is between 9 and 20, and between a third and a half of them exist because of a decision made two years earlier by someone who has left. Data the product could ask for later. A verification email that could have been a magic link. An onboarding tour standing in front of the thing the tour is about.

Cut half. Then do the walkthrough again, because removing steps reveals the ones that were hiding behind them.

What goes wrong here. The cut gets negotiated down to the two steps nobody owns, and the fields sales wants stay because sales wants them. The way through that is to move those fields rather than delete them: ask after the first value, not before it, where the answer rate is higher anyway because the person now has a reason to answer.

Step 4: Put the first value in the first session

A 14-day trial is not 14 days of consideration. It is one session, and then a decaying chance of a second one. If the action from step 2 happens on day 6 for your median payer, your real conversion ceiling is the share of people who come back on day 6.

So bring it forward. Pre-fill the account with sample data that looks like theirs, and ship the one template that covers the common case. If the product truly cannot show value without their own data, then the first session's only job is to get that data in, and everything else on the screen is in the way.

What goes wrong here. The team builds a better empty state instead of removing the emptiness. An illustration and a cheerful line is still an empty product. Sample data a person can delete in one click beats any explanation of what the product would do if they had data.

Step 5: Write five messages for five decisions, not for five days

Most trial sequences are scheduled: day 1, day 3, day 7, day 11, day 13, each a little more urgent than the last. Rewrite them against the five decisions a person actually makes.

1. Is this the kind of tool I thought it was?

2. Can I get it to do the one thing I came for?

3. Will this work with the systems I already have?

4. What happens to my work if I stop paying?

5. Can I defend this price to the person who approves it?

One message per decision, triggered by behavior rather than by date, and the one for a decision already made does not send: somebody who has connected their systems does not need message 3. The fifth is the one almost nobody writes, and in anything sold to a team it decides the trial, because the person in the trial is rarely the person with the card. Give them the numbers in a format they can forward.

What goes wrong here. The sequence gets louder instead of more specific. Three "your trial ends tomorrow" emails raise unsubscribes and move the paid rate by nothing I have ever been able to measure. Urgency works on a decision already made. It does not make a decision.

Step 6: Re-cut the rate before you celebrate

A single trial-to-paid number averages at least three populations that behave nothing alike: people who came from search knowing what they wanted, people who came from a paid ad and are still deciding, and people a salesperson sent. Split by source, by company size and by whether a human spoke to them, then look at the lowest cell. In most companies one channel is quietly producing most of the trials and almost none of the customers, and the blended rate hides it. That cell is a budget question rather than a product one, and it belongs in cost per qualified lead.

What goes wrong here. The rate improves, the team claims the change, and the real cause was a shift in traffic mix. Compare within each source, not across the blend.

Should you ask for a credit card?

Trials that ask for one converted at 30 percent in the SaaS Conversion Report, more than 5 times those that don't, and only 20 percent of products ask. The card persuades nobody: it removes the people who were never going to buy before they reach the denominator, so the rate rises while new customers may fall. Keep a holdout, and read the fuller version in revenue growth.