Should you optimise Meta ads for trial starts or purchases?

You are setting up a Meta campaign for your subscription app and you hit the dropdown that decides everything. Optimise for trial starts, or optimise for purchases. Pick trials and the campaign will fill up fast. Pick purchases and it might barely spend. Everyone tells you something different, so you pick trials, because at least something happens.

Here is the short version. Trial starts are the right call when you are starting out, and the wrong call the moment you can afford to leave them. But most founders never actually make this choice. They optimise for trial starts because that is the only event Meta is receiving, and they find out months later that the decision was made for them by a tracking gap, not by strategy.

What each event actually tells Meta

Meta is not evaluating your business. It is a matching engine, and whatever event you choose becomes the only definition of success it has.

Optimise for trial starts and you are telling Meta: go find people who will tap the button and begin a free trial. That is what it will do, with impressive efficiency. It will learn what a trial starter looks like, then go find thousands more of them.

Optimise for purchases and you are telling Meta: go find people who will pay. Different target, different audience, usually a more expensive one, and a much smaller pool of signal to learn from.

Neither instruction is wrong. But notice that the first one says nothing at all about money. Meta does not know that a trial is supposed to become a subscription. It only knows the finish line you drew.

The case for trial starts

The argument for trials is volume, and it is a real argument.

Meta needs roughly 50 conversion events per ad set per week to get out of the learning phase and start optimising properly. Below that, delivery stays unstable, costs swing wildly, and you cannot tell a bad creative from a bad day. If you are spending 50 or 100 dollars a day on a subscription app, you will hit 50 trial starts a week long before you hit 50 purchases.

So starting on trial starts is usually correct. You get the campaign out of learning, you get enough data to judge creative, and you get a functioning account instead of one that stutters. Anyone who tells you to optimise for purchases on day one at a small budget is describing a campaign that will never exit learning.

The mistake is treating that starting point as the permanent setup.

The trap: Meta will find you trials that never convert

This is the part that catches people, and it is worse now than it used to be.

Trial start optimisation is blind to everything that happens after the trial begins. Meta has no idea whether your trial starters convert at 40 percent or 4 percent. It just knows they started trials, so it keeps hunting for more people who look like the ones who started trials.

The people most likely to start a free trial are not the people most likely to pay for it. They skew younger, they skew more price sensitive, and a meaningful share of them are trial tourists who cancel inside the first 48 hours. Meta will find them for you, cheaply and at scale, because you asked it to.

Hand this setup to Advantage+ and it gets sharper still. Broad automated targeting is very good at finding the cheapest version of whatever you asked for. Ask for cheap trials and you will get remarkably cheap trials from exactly the cohort least likely to ever pay you.

The pattern looks like this. Cost per trial drops week over week. You scale spend because the dashboard says it is working. Then a month later, when those trials should have converted, revenue has not moved. Nothing in Ads Manager tells you this happened, because Ads Manager is still reporting on the metric you told it to care about, and that metric looks great.

Most people never actually made this choice

Here is the part that gets skipped in every guide I have read on this.

You cannot optimise for purchases if Meta is not receiving purchase events. And by default, it is not. Trial conversions and renewals fire server side, days or weeks after the install, long after the phone stopped talking to the Meta SDK. Unless you have deliberately wired your real paid events into Meta through the Conversions API, the purchase option in that dropdown is either missing or attached to an event with almost no volume.

So the founder opens the dropdown, sees that purchases would give them a handful of events a week, picks trial starts, and calls it a decision. It was not a decision. It was a tracking gap making the call for them.

This is why the sequencing matters. Fixing your event pipeline is not a reporting nicety you get to later. It is the thing that gives you the option to optimise for money at all. The RevenueCat to Meta Conversions API setup is the specific work involved, and if you have already done it but the numbers look nothing alike, that gap has its own explanation.

How to know you are ready to switch

Do not switch on a feeling. Switch when the numbers say the campaign can survive it.

You are sending real purchase events. Trial conversions and initial purchases are reaching Meta through the Conversions API and you have verified them landing in Events Manager, not just assumed it.

You have about 50 purchase events per ad set per week. Not per campaign, per ad set. If you are close but not there, consolidate ad sets rather than staying on trials.

You know your trial to paid conversion rate. If you do not know this number, you cannot evaluate either setup, because cost per trial is meaningless without it. Cost per trial divided by conversion rate is your real cost per subscriber, and that is the number to compare against what a subscriber is worth.

Your cost per paying subscriber is below your LTV with room to spare. Not equal to it, below it, with margin for refunds and churn. You can check where that line sits in a couple of minutes with the LTV to CAC calculator, and model the full install to profit path with the app install profitability calculator.

When those four are true, move to purchase optimisation and expect your cost per event to look worse on the dashboard while your revenue looks better. That trade is the whole point.

If you cannot get to 50 purchases a week yet

Most apps at small budgets cannot, and that is fine. You have better options than just waiting.

Stay on trial starts, but stop judging the campaign on cost per trial. Track cost per trial alongside your trial to paid rate every week, and treat the product of those two as your actual KPI. The moment cost per trial falls while trial to paid falls faster, Meta has started buying you the wrong people, and you should tighten targeting or change the creative angle rather than celebrating.

Turn off Advantage+ campaign budget while you are on trial optimisation, or at minimum run it against a manual ad set so you can see the difference in downstream conversion, not just in cost per trial.

And if your paywall gates the trial behind a card, your trial starts are already a much stronger signal than an app where anyone can start one with a tap. Card-gated trials convert far better, which means trial optimisation stays honest for longer. Ungated trials are where this goes wrong fastest.

Bottom line

Optimise for trial starts to get the account learning, then move to purchases as soon as your event volume allows it. That is the simple version, and it is right.

The version that actually matters is this. Trial start optimisation tells Meta to find people who start trials, and Meta will do that job so well that it will find you the people who start trials and never pay. Purchase optimisation is the only setting that points Meta at your revenue, and you cannot use it unless your real paid events are reaching Meta in the first place. So the dropdown is not really the decision. The decision is whether you fixed your tracking, because that is what determines which options you get to pick from.

If you are running a subscription app on trial optimisation and you cannot tell whether those trials are turning into a business, that is exactly the kind of thing I pull apart on a teardown call. I will look at what events are actually reaching Meta, what your true cost per paying subscriber is once the trial rate is accounted for, and whether your campaigns are chasing subscribers or just cheap sign-ups.