How Long Should Your Free Trial Be? Start With Time to Value

Someone asked me this last week and I could feel the argument loading up. Seven days or fourteen? An investor had told them thirty. Their closest competitor runs fourteen with a card up front. They had four opinions, no data, and they wanted me to pick a side.

I didn't pick one, because the question is broken. You are not choosing a number. You are choosing how much runway to give someone to reach the moment your product clicks. Sort that out and the number falls out on its own.

Here is the short version, then the math underneath it:

Why is "how long should my free trial be?" the wrong question?

Because a trial length in isolation carries no information. Fourteen days is luxurious for a product someone gets a result from in an afternoon, and brutal for one that needs a data import plus a teammate's permission. Same number, opposite outcomes.

What people are really asking is how long someone needs to decide. The honest answer is that nobody decides on a calendar. They decide at the moment the product either proves itself or doesn't. Everything after that point is paperwork, and everything before it is suspense.

Nobody converts because a trial ended. They convert because the product proved itself, and the trial happened to still be running when it did.

So the number you need isn't the trial length. It's the time it takes to reach the proof.

What is time to value, and why does it set your trial length?

Time to value is the gap between signup and the first moment the product does the thing the customer showed up for. Not first login. Not "completed onboarding". Not a checklist with green ticks. The moment they get an actual result: the report that answers their question, the first invoice sent, the first teammate replying inside the tool. That's the aha moment, and it's the only event in the whole funnel that predicts whether they stay.

Your trial has exactly one job: stay open long enough for a typical person to get there, with enough slack left over to make the call. Longer than that adds nothing. Shorter than that kills people who would have paid you, which is the expensive kind of churn because you never even see it. This is the point Andrew Chen has been making for years about retention being set upstream of everything else, and it's why his power user curve is more useful than a single conversion rate. Reforge frames it the same way: early engagement is the leading indicator, not the lagging one.

Here's what that looks like across three real product shapes. Same question, three completely different answers.

Look at the middle bar. Four days to value, fourteen day trial. That extra window isn't padding, it's the room a real person needs to hit a weekend, get pulled into something else, come back, show a colleague, and then decide. Squeeze it and you lose people who liked the product but ran out of clock.

How long should your free trial actually be?

Here's the formula I use, and it has two terms:

Trial length = the day by which most of your would-be activators have hit value + a decision buffer.

The first term is not your median. This is where people get it wrong. Time to value has a long right tail: a chunk of users get there on day one, most get there in the first few days, and a slow group trickles in for a fortnight. That shape is close to an exponential distribution in most products I've looked at, sometimes closer to log-normal, and the practical consequence is the same either way: covering 85% of activators takes roughly two and a half times your median, not one times it.

The second term is the human bit. Somebody has to notice they got value, decide it's worth paying for, and possibly ask a manager. Two to four days for self-serve, longer if there's a budget holder involved.

Put your own numbers in:

Where this number comes from: it's a coverage calculation, not a model of buying behaviour. If your median time to value is M days and activation follows the usual long-tailed shape, the day by which a share X of activators have arrived is M multiplied by log base 2 of 1/(1-X). At a 3 day median that puts 80% coverage at day 7, 85% at day 8, and 90% at day 10, which is why the jump from 14 to 30 days buys so much less than it looks like it should. Add your decision buffer and you have the trial length. Two caveats worth stating plainly. First, measure the median on people who converted and stuck around, not on all signups, because the signup pool is full of tourists who were never going to activate and they'll drag your median toward infinity. Second, if you already know your real 85th percentile from a survival curve, use that number directly and skip the estimate. The maths here is a stand-in for data you may already have.

What does a trial that's too long actually cost you?

Most teams treat a longer trial as harmless generosity. It isn't. Every extra day past the decision point costs you something specific.

The first cost is attention. A trial with weeks left on it is not urgent, and Parkinson's law applies to evaluations as much as to work. People plan to look at it properly next week, and next week they've forgotten the password. The unfinished-task pull that keeps someone coming back only exists while the task feels live.

The second is cash and learning speed. A 30-day trial means you find out whether your funnel works a month at a time. Halve the trial and you double your iteration rate on onboarding, pricing, and emails, and you pull revenue forward, which moves your CAC payback period in the direction you want. Payback and churn are on a16z's list of the metrics that actually constrain a business for a reason, and David Skok's work on why churn is critical shows how quickly slow cycles compound against you.

A 30-day trial doesn't give people 30 days of consideration. It gives them 29 days of forgetting and one day of panic.

The third cost is the one that stings: a long trial hides your activation problem. If people are taking eleven days to hit value, that's worth knowing and fixing. Stretching the trial to accommodate it just buries the signal under a comfortable number. Lenny Rachitsky's retention benchmarks and Amplitude's work on retention curves both point the same direction: the early days decide the outcome, so that's where the fixing belongs.

So is 7, 14, or 30 days right for you?

Match the row to your measured time to value rather than to your industry or your competitor.

One thing worth saying about the card-up-front question, since it always comes up alongside length. Requiring a card typically cuts trial starts by half and roughly doubles the conversion rate of the ones who start, so the paying-customer count often lands in a similar place. Stripe supports both patterns without much extra work, so treat it as an experiment rather than a philosophy. Recurly's subscription research is a decent sanity check on what normal looks like in your category.

How do you measure your own time to value?

This is an afternoon of work and one SQL query, and almost nobody does it. Five steps.

1. Define one value event, precisely

Not signup, not "finished onboarding", not a vanity milestone. Pick the single event that means the product did its job: first report shared, first payment collected, first ticket resolved through the tool. If your team can't agree on one, that disagreement is your real problem and it's worth an hour to settle. ProductLed and Lincoln Murphy's writing on customer success are both good on choosing this event honestly rather than flatteringly.

2. Only look at people who converted and stayed

Filter to customers who paid and were still around at 90 days. Their path is the one you want to reproduce. Including everyone drags your median toward people who were never going to activate. Do watch for survivorship bias here: you're deliberately studying survivors, which is fine for setting a trial window, and misleading if you start treating it as the average user's experience.

3. Pull the hours between signup and that event

Hours, not days. Days round away the difference between "same afternoon" and "next morning", and that difference matters a lot at the short end. Group by signup week so you can see whether the number is moving.

4. Take the median and the 85th percentile

The median tells you what typical looks like. The 85th percentile tells you how long the tail runs, and the tail is what your trial has to cover. If your median is 2 days and your 85th percentile is 16, you don't have one product experience, you have two, and they probably map to two different segments worth separating.

5. Compare it against your current trial

Now the number means something. If your trial ends well after the 85th percentile, you're paying for dead air. If it ends before it, you're cutting off customers mid-evaluation. A cohort retention chart alongside this makes it obvious which one you've got.

What if your time to value is genuinely long?

Then a longer trial is the wrong fix. It's the fix that treats the symptom.

If it takes eleven days to get someone to first value, the work is shortening those eleven days, not extending the window around them. Ship sample data so the product isn't empty on day one. Ship templates so nobody stares at a blank canvas. Do the integration for them on a call. Cut the setup steps that don't gate the value event. The activation milestones playbook is the version of this I run most often, and it usually pulls days out of the front of the funnel rather than adding them to the back.

There is a real exception. Some products have a structurally long evaluation: security review, procurement, a data migration that needs someone else's engineering time. If that's you, stop calling it a trial. Call it a pilot, put a human on it, and give it a schedule with checkpoints. An unattended 60-day trial for an enterprise buyer is just a slow way to get ghosted, and Bessemer's benchmarks on scaling show how much of that outcome is decided by process rather than product. First Round Review has good field reports on running pilots that don't drift.

Should you use a reverse trial instead of a fixed one?

A reverse trial hands new signups the full paid product for a short window, then drops them to a permanently free plan rather than locking them out. I like it when time to value is unpredictable, because the slow users aren't punished for being slow, and the downgrade moment does the selling for you. Losing a feature you've started relying on hits harder than never having had it, which is loss aversion doing useful work.

It only holds up if the free tier is limited enough that the drop actually stings. That's the same trap I wrote about in whether you should have a free tier at all: a free plan generous enough to solve the whole job stops being a funnel and starts being your product. Brian Balfour's work on model and channel fit is the sharpest thinking I've found on when this pattern earns its keep.

The part that matters more than the number

Once you've set the length correctly, the biggest remaining lever isn't the length at all. It's what happens at the end of it.

Most trials die in silence. The product stops working one morning and the only communication is a billing page nobody asked for. That moment converts far better when the email references what the person actually built during the trial instead of counting down days at them, and better still when someone who was clearly mid-setup gets offered a short extension rather than a wall. I've written the sequence up in how to reduce trial expiry abandonment, and the trial expiry email experiment has the exact timings. The free-to-paid abandonment playbook covers the checkout side of it.

It's also worth being honest about what a trial can and can't fix. If people reach value and still don't pay, the problem is your offer or your price, which is a different piece of work entirely and one I covered in churn as an offer problem. If they convert and then leave in month two, look at first-month retention instead, because that's where the real answer lives. HBR's research on customer effort and Nielsen Norman Group's usability work both land on the same conclusion: reducing friction beats adding delight, and a trial is mostly a friction test.

Stop asking how long the trial should be. Ask how fast someone can get a result, then give them that plus a few days to think.

If you want to know whether your trial length is even in your top three problems, the free 60-second churn health check asks a handful of questions and tells you where your biggest leak actually is, plus the next three things to fix. And if you want to see what the compounding looks like once those trial customers become paying ones, the MRR churn impact simulator runs the numbers over 24 months.