I removed the free trial for paid Meta traffic on a subscription app. Conversion jumped from 7.1% to 12% and cost per result fell from $58 to $40, and every one of those results was a purchase. A result meant a trial or a purchase before the switch and only a purchase after it, so the two costs are not the same unit.
By early June, a result on this app cost me $58 in ad spend, and since a result could be a trial or a purchase, I was paying full price for a maybe.
So I cut the trial. The next week, conversion jumped from 7.1% to 12% and cost per result dropped from $58 to $40. I started getting pure purchases and stopped getting trials, which had been the majority of the events.
The setup
A subscription app, paid Meta traffic, campaigns optimized for purchase from day one. The paywall offered two options: a weekly purchase, or a yearly plan behind a free trial.
Those are two different users. The weekly buyer already decided to pay. The yearly trial user made a promise to decide later, and for this app, later mostly never came. For some apps, having a trial is a big risk, as trial users may not be as motivated or have plenty of options to choose from, especially coming from paid sources. Think about generic terms in Apple Search Ads, or someone looking for a specific app who sees different ones on Meta or TikTok.
The ad account treated both events as a win, naturally. Meta, as of 2026, can’t really optimize for more than one event properly (yes, there are workarounds). Trial starts and purchases both counted as results, both fed the optimization, and both cost about the same to buy. But trials converted poorly for this app.
One variable moved
On June 9, 2026, I removed the trial for paid Meta traffic. Everything else stayed where it was:
- Same campaign, untouched
- Same creatives
- Same purchase optimization event
- Same prices, plans and paywall design, and no app release
Because the change lived on the paywall and not in the ad account, Ads Manager never put the campaign back into learning. That is as clean as a before and after gets. It is still a before and after. On June 9 the event changed meaning, from trial or purchase to purchase only, and nothing held delivery constant.
What happened
Conversion had been climbing for four weeks with my older model. Fine. 3.4%, 4.8%, 4.9%, 7.1%. The week the trial came off, it hit 12%.
One Meta campaign, results divided by installs. June 9 is the first week without the trial.
“Samet, that could just be momentum.”
Fair. The account was maturing and the rate was already trending up. But the biggest jump landed exactly on the switch, and it came with the part that surprised me more. And remember that the trials were barely converting for this app.
Meanwhile, naturally, cost per result kept falling. Removing the trial removes a whole class of conversion events, and fewer events usually means the algorithm struggles and costs climb. Instead, cost per result went from $58 to $40, and every $40 result was a purchase.
Every week, no cherry picking
Most case studies show you the two weeks that make the author look good. So I’m showing the whole run, including the early weeks where I paid $243 per result like a tourist.
Same campaign and same optimization event throughout. Until June 8 a result was a trial or a purchase. From June 9 only purchases could fire it, so $40 is the cost of a purchase.
Why it made sense
The economics. This was never a cost play. By the final trial week, a trial cost about what a purchase cost. The usual argument for keeping trials on paid traffic is that they give you cheaper volume to feed the algorithm. With a result at $58 and trials no cheaper than purchases, that argument was gone.
The signal. Meta builds your audience around whoever fires your optimization event. When trial starts and purchases both fire, you train the system on a mix of committed buyers and maybes. You can never get what you want, unless you pass placeholder values and optimize for value. Remove the trial and every event is a payment from the kind of user you want more of. People worry that optimizing for purchases alone starves the algorithm, but here it didn’t. 54 purchase events in a week kept this campaign just above the 50 events a week line I want every campaign to clear.
The traffic. A user who reaches your paywall from an ad has already passed three filters: they stopped scrolling, they tapped, they sat through onboarding. The ad did the convincing. At that point, my read is that the trial mostly gave a decided user a reason to postpone.
Do trials still make sense?
Sometimes. A trial is a tool, and most apps run it as a default setting.
In RevenueCat’s latest State of Subscription Apps, hard paywall apps convert about five times better than freemium: 10.7% versus 2.1%. Retention of annual subscribers in year one is nearly the same, 27% and 28%. Revenue per install at day 60 is about eight times higher, $3.09 versus $0.38. Their split is freemium versus hard paywall, a different lever than trial versus no trial, so it is context for this test, not proof. Asking for commitment earlier doesn’t wreck the funnel.
The same report shows where trials work. Trials of 17 to 32 days convert at about a 42% median, roughly 70% better than trials of four days or less. Some products need a week or two of use before the value shows, like a meditation app or a language app. I’ve also seen apps that convert 50%+ of their trial users (not surprisingly, an established app with no competitor). Well done to them.
So before you copy this test, ask two things. Does your product need time to prove its value? And does your rate from trial to paid show that it does? A strong rate means the trial earns its place, while a weak rate means a leak, and on paid traffic you fund that leak with your own budget.
The playbook
If you run paid traffic to a trial paywall, I would test it like this:
- Remove the trial for paid traffic only. Paid users already showed intent. See whether a cleaner purchase signal helps.
- Keep the trial for organic users as a sanity check, not a control. Watch that conversion rate. If free trials won’t convert even organic users, that points at the product, not the ads. If your paywall tool can split paid users at random between the two offers, use that split as your control instead.
- Change the paywall, not the ad account. Same campaign, same creatives, same event. The campaign stays out of learning and fewer things move at once.
- Give it a full week. iOS attribution needs a day or two to settle. Don’t judge it on day one.
- Watch what happens after the purchase. Refund rate and early churn on short plans tell you whether these buyers stick.
The caveats
This is one app, five weeks, and a few hundred installs per week. I’m deliberately not naming the app, the geo, or the account structure. Conversion was already climbing before the switch, there was no holdout, and a result meant something different after June 9, so $58 and $40 are not the same unit. The mechanism makes sense, but it’s still one test. Run it on your own numbers before you believe it.
The bottom line
When a trial costs what a result costs and rarely converts, you’re buying maybes at full price. For this app, the maybes were the most expensive thing in the account.
If you test this on your own traffic, tell me what happened. My favorite messages are the ones with a screenshot of the before and the after.