How Free Trials Quietly Break Your CAC and LTV Math
The number that got me was 82%. That's roughly the share of opt-in free trial signups that never pay a cent, if you take First Page Sage's 2025 benchmark of an 18.2% trial-to-paid rate at face value. And yet a spreadsheet I inherited at one company was cheerfully dividing marketing spend across every one of those signups and calling the result CAC. The CAC looked fantastic. It was fiction.
If you count free trial signups as customers, you understate CAC and overstate conversion at the same time, which is the worst possible pair of errors to make together: it tells you to spend more on a channel that's actually underwater. The fix is boring and it's two parts. Define "customer" as a paid conversion, not a signup. Then start the revenue clock when the money starts, not when the trial does.
Let me show the damage before I show the fix, because the damage is bigger than most people expect.
The two-signup trap
Here's the napkin version. Round numbers on purpose.
Say you spent $100,000 last quarter on marketing and it produced 5,000 trial signups. Of those, using that 18% figure, about 900 convert to paid.
Divide by signups and you get a CAC of $20. Divide by actual paying customers and you get $111. Same spend, same quarter, a 5.5x gap depending on which denominator you grabbed. I have watched a growth team greenlight a channel on the $20 number and then spend six months wondering why the payback curve never showed up. It never showed up because the $20 wasn't real.
Paddle's team puts the principle plainly: free or trial users carry no guaranteed monetary value, so you shouldn't fold them into your count of acquired customers when you compute CAC. Interested is not acquired. A trial is a maybe.
The reason this trap is so easy to fall into is that "signup" is the event your tooling fires most loudly. It's the top of the funnel, it's instrumented to death, it's the number in the daily Slack digest. Paid conversion happens quietly, sometimes weeks later, often in a billing system that doesn't talk to your marketing dashboard. So the convenient number wins, and the convenient number is wrong.
And it's not a niche problem. ChartMogul's 2025 conversion report pegs free trials at around 57% of SaaS companies versus 26% on pure freemium, so most of the market is running exactly the motion where this error lives.
The corrected CAC formula
The correction itself is not clever. It's discipline about what goes in the denominator.
| Version | Formula | Same example |
|---|---|---|
| The wrong one | Marketing & sales spend ÷ all trial signups | $100,000 ÷ 5,000 = $20 |
| The honest one | Marketing & sales spend ÷ paying customers acquired | $100,000 ÷ 900 = $111 |
| Fully loaded | (Spend + trial serving cost) ÷ paying customers acquired | ($100,000 + trial hosting/support) ÷ 900 = $111+ |
That third row is the one people skip. Every trial you run has a cost to serve, even the ones that never pay: hosting, support tickets, the human who answers "how do I export this." Those costs are part of acquiring the customers who do convert. You can fold trial serving cost into CAC or treat it as a churn-of-free-users line item, but you have to put it somewhere. Pretending free trials are free is how a "profitable" motion runs a quiet deficit.
One assumption I'm making, and you should feel free to argue with it: I'm attributing the full quarter's spend to the customers who converted that same quarter. In a fast trial (14 days, monthly plans) that's roughly fine. If your trial is 30 days and your sales cycle is 90, you've got a cohort-timing problem, and you should be matching last period's spend to this period's conversions. More on timing next, because it's the second half of the bug.
Now the LTV side, where it gets worse
CAC is only half the ratio. The trial distortion sneaks into LTV too, and it does it through timing.
LTV, at its simplest, is (average revenue per customer × gross margin) ÷ churn rate. Two of those three inputs get quietly poisoned by trials if you're not careful.
First, margin. Improvado's 2026 LTV:CAC guide makes the point that LTV has to run on gross margin, not gross revenue, or the ratio comes out inflated. A customer paying $100 a month at 80% margin contributes $80, not $100. Fine, that's the standard warning. But trials add a twist: the free period has negative margin. You're serving product and collecting nothing. If your LTV model starts the revenue clock at trial signup, you're booking value during a window where the customer is costing you money and paying you zero.
Second, the customer lifetime itself. If you define the lifetime as "months since signup," every customer's clock starts weeks before they pay. Your average lifetime looks longer than the paying relationship actually is, so LTV inflates. It sounds small. Over a big cohort it isn't.
The clean fix is a deferred-revenue timing correction, and it's one sentence: the LTV clock starts at first payment, and the free period is booked as acquisition cost, not as month zero of the relationship. Orb's writeup on CAC payback calls the lag between acquisition spend and deferred revenue recovery the defining structural feature of SaaS economics, and notes the median payback sits around 16 months. A trial just widens that lag. Model it honestly and your payback period gets longer and truer; model it by signup date and you get a payback curve that pays back before the customer has paid you anything, which is nonsense you can't take to a CFO.
A corrected worked example
Let me redo the napkin math end to end, with the timing fix in.
- Spend: $100,000. Trial signups: 5,000. Paying conversions: 900.
- Honest CAC: $111 per paying customer.
- Each paid customer: $100/month, 80% gross margin, so $80/month contribution.
- Churn: 4% monthly, giving an average paying lifetime of 25 months.
- LTV on gross margin, from first payment: $80 × 25 = $2,000.
- LTV:CAC = 2,000 ÷ 111 ≈ 18:1.
Now watch what the signup-counted version would have told you. CAC of $20, and an LTV that starts a month early so the lifetime reads 26 months of revenue including a free one, call it $2,080. Ratio: 104:1. A CFO I've sat across from would take one look at 104:1 and ask why on earth you aren't spending ten times more on marketing. You'd oblige. You'd pour money into a channel whose real ratio was healthy but ordinary, on the strength of a number that was five times too flattering, and you'd find out at the payback checkpoint, which is the most expensive place to find out.
Even 18:1 there is on the rosy side because I used round churn and margin to keep the arithmetic clean; the point isn't the exact figure, it's the size of the gap between the two methods.
Where the good conversion numbers actually come from
One more trap, because it feeds the others. The trial model changes the conversion rate so drastically that a blended benchmark is close to meaningless. Opt-in trials, no card required, run around that 18.2% First Page Sage figure. Opt-out trials, card up front, convert near 48.8% in 1Capture's 2025 look at more than 10,000 SaaS companies. That's not a small spread. It's the difference between one in five and one in two.
So if you benchmarked your card-required trial against an industry-average conversion rate, you'd think you were failing when you were fine, or the reverse. Segment your conversion rate by trial type before you compare it to anything. And when you compute CAC, remember the opt-out trial has already filtered out most tire-kickers, which is exactly why its denominator behaves better.
The instrumentation problem underneath all of this
Here's the thing nobody wants to say: this is rarely a math failure. It's a plumbing failure. The signup event lives in your product analytics, the payment event lives in Stripe or your billing system, and the ad spend lives in a third place, and no single tool sees all three cleanly. So the denominator defaults to whatever's easiest to query, which is signups.
The durable fix is to define the paid-conversion event as your "customer created" moment everywhere, and to gate every LTV cohort on that event rather than on signup. Most product analytics tools, Kixo among them, let you build cohorts on a specific paid-conversion or activation event instead of first-touch, which is what keeps trials out of the customer count in the first place. Whatever you use, the rule is the same: one event defines a customer, and it's the one attached to money.
Two things worth wiring up alongside that. If your trial is really an activation race, you want to know how fast trial users hit first value, because that's what predicts conversion better than the signup count ever will. That's the time-to-value measurement problem, and it pairs directly with finding your product's aha moment so you can tell early which trials will convert. Get those two instrumented and your CAC denominator stops being a guess.
If you want the ratio to live somewhere your finance team trusts, put it on the same surface as payback and margin. I keep mine in a unit economics dashboard so the CAC denominator and the LTV clock are visible in one place, which makes it a lot harder for the convenient-but-wrong number to sneak back in.
Quick answers
Should free trial users count in CAC at all? Not as acquired customers in the denominator. Their serving cost can go into the numerator as part of what it took to win the ones who paid. Interested isn't acquired.
When does the LTV clock start? At first payment. The free period is acquisition cost, not month one of the relationship. Starting at signup inflates both lifetime and revenue.
Why does my LTV:CAC look too good? Usually one of two things: you counted signups as customers, shrinking CAC, or you started LTV at signup and used gross revenue instead of gross margin. Often both at once, which is how you get a ratio in the triple digits.
Does the trial type really change the answer? Yes. Opt-in and opt-out trials convert at wildly different rates, so segment before you benchmark and before you set the denominator.
The whole correction is two habits, and neither is hard. Count paying customers, not signups. Start the revenue clock at the first payment. Do those and your unit economics get less exciting and a great deal more true, which, the one time I actually did it cleanly, was the thing that let me defend a budget instead of quietly hoping nobody checked the math.