Your Meta dashboard shows 220 trials at $9 each. Your product dashboard shows that four became paying accounts. I have seen this split more times than I care to count: Meta is finding cheap signups because that is what the campaign asks it to find. This list is for SaaS founders and growth leads whose trial volume looks fine while activation stays flat. Each fix moves the signal closer to someone who uses the product. Cost per trial may rise. That can be the healthy result.

1. Stop asking Meta to buy signups

Meta does not judge the quality of a trial. It buys the event you point it at, as cheaply as it can. If that event is trial start, people who like starting trials look valuable to the algorithm whether they connect a data source, run a report or never log in again. The dashboard calls them conversions. Your product has a less generous definition.

I used to tell clients to optimise for trial start because volume helps delivery. I was wrong to treat volume as the goal. In one self-serve comparison, a set producing 220 trials at $9 yielded four paid accounts. Another produced 140 trials at $15 and yielded 19 paid accounts. The second set looked worse on cost per trial and much better on the outcome that pays the bills.

First step: define the first action that proves a user reached value. That might be connecting a data source and creating one report; use the moment that fits your product, not a generic page view. If you have enough of those events to support delivery, test activation as the optimisation event. Judge the test on cost per activated trial, then paid conversion. Give it two weeks rather than declaring the higher trial cost a failure on day two.

2. Fire a real signup event, not a thank-you-page view

A Pixel PageView on /thanks tells Meta that a page loaded. It does not prove a user exists. Reloads can inflate the count; a curious visitor who never confirms an email can still look like a successful trial. If you train delivery on that event, you are asking Meta to find more page loaders. It may oblige.

First step: compare reported trial conversions with validated user rows in your database. Then fire Lead from the server when a real signup passes validation, and keep the thank-you-page view for diagnosis rather than optimisation. A separate event can record activation when the user completes the action you defined in No. 1. The point of the two-event setup is not tidier naming. It is to distinguish someone who entered an email from someone who used the product.

Expect your reported trial count to change if the old event included page loads that were never signups. That is not lost demand. It is a measurement problem you can finally see. Fix the count before you ask an algorithm to learn from it.

3. Send activation back through Conversions API

A clean signup event still leaves Meta guessing about quality if the useful action happens later in your app. The browser Pixel can see a visit and miss the account that activates two days later. Meta keeps buying people who resemble signups because it never hears which signups mattered.

Conversions API for CRM gives you a way to send later stages back. Meta reports about 21% lower cost per quality lead on instant forms and 9.5% lower on website forms for advertisers using this approach. Those figures are a reason to examine your signal, not a forecast for your account. The mechanism is straightforward: delivery can learn from a stage closer to a useful customer instead of treating every email address alike.

First step: send one activation event, such as ActivatedTrial, with the matching information needed to connect it to the signup. Check that the event fires for the right users before you optimise against it. If your paid conversion arrives too late or too rarely to guide delivery, use an earlier qualified stage that enough accounts actually reach. Keep paid conversion in the report. A middle-funnel signal is a bridge to that result, not a substitute for measuring it.

4. Make the ad sell the job, not the word “free”

If your headline shouts Free Trial, No Card and Set Up in 30 Seconds, it filters for people who like low-commitment offers. Some will be buyers. Others collect free tools the way the rest of us collect open browser tabs. Meta sees who responds, then looks for more people like them.

Creative does targeting before the targeting controls get a vote. An ad that names the painful job and shows the screen where that job gets done gives the right prospect something to recognise. It also gives the wrong prospect a reason to keep scrolling. Fewer clicks and a higher CPC can be a fair trade if the people who arrive actually use the product.

First step: take your highest-spend trial ad and rewrite its opening. State who the product is for and what changes for that person in week one; move the free-trial offer to the caption or the end. Keep the old ad as a comparison. Do not crown the new one because its click-through rate looks respectable. Follow both groups through activation and paid conversion. The useful ad is the one that attracts people with the problem, not the one that makes a free button irresistible.

5. Send trial prospects to a signup that asks something of them

Pre-filled Instant Forms remove effort. That is useful when you want an easy response and a poor filter when your product itself requires work. A person who would have left after reading your pricing page can become a cheap form completion instead. Meta records the conversion; your team inherits the curiosity click.

For trial and demo campaigns, practitioners favour sending prospects to the site, where they can see what the product asks of them before signing up. I would rather pay $22 for someone who typed a work email, picked a use case and saw the integration list than $8 for someone who tapped twice without understanding the job ahead. Those are illustrative prices, not a promise that more friction automatically creates a buyer.

First step: compare Instant Form trials with on-site trials on activation, not form completion. If the native form fills your dashboard while the site fills your product, move the trial ask to your domain. Ask for information that helps a prospect assess fit; do not add fields merely to make the form feel serious. Useful friction reveals intent. Pointless friction just loses it.

6. Build lookalikes from people who use and pay for the product

A lookalike reflects its seed. Feed it everyone who has ever opened a free trial and it looks for more people who resemble free-trial starters. Feed it active, paying customers and you give it a different pattern to follow. Seed quality matters more than a large list of people who never made it past signup.

Teams running SaaS campaigns recommend high-LTV active customers over raw signups. A 1% lookalike seeded from buyers is one way to begin when you have the customer data for it. It is not a magic setting that repairs a weak seed.

First step: create separate audiences from activated accounts and paying customers, then compare them with the all-signups audience you have been scaling. Keep the distinction visible in your reporting. If the all-signups audience produces cheaper trials but fewer activated users per dollar, stop calling it the winner. Choose the seed for the customer you want more of.

7. Grade creative on activation, not CPC

I have killed ads because they looked expensive at the top of the funnel. A product screenshot at $4.10 CPC loses a click contest to a meme at $0.90 CPC. But the screenshot may turn away people who do not have the problem, while the meme invites them in. The cheap click is only cheap until you pay for enough of them to find one customer.

The same problem shows up in trial cost. Two creatives can look alike on cost per demo and differ two to three times on cost per retained customer because they attract different people. The chain matters: click, trial, activation, paid account, retention past payback. Stopping at the first inexpensive link makes the report prettier and the decision worse.

First step: add activated trials per $1,000 spent to your creative test sheet, beside cost per trial and paid conversion. Keep CPC as a diagnostic for what happened at the click; do not use it as the verdict. Wait until an ad has enough activation data to make a comparison worth acting on. If a screenshot costs more to click but brings in users who do the work your product requires, let the screenshot be boring. It has a job.

8. Make expansion earn its place in a niche campaign

When expansion treats your audience as a starting suggestion, delivery can move toward people who convert more cheaply. For a niche B2B SaaS product, cheaper trial starters outside your intended market may be exactly the wrong people. The algorithm cannot infer your fit requirements from a signup event you have not connected to activation.

This does not mean deleting Advantage+ because it has a suspiciously cheerful name. It means refusing to judge expansion on signup cost alone. First step: compare delivery with and without expansion in trial campaigns, and watch cost per activated trial alongside the audience it reaches. Give the comparison time to settle before opening things further. If expansion cuts cost per signup by 30% while the activation rate halves, you did not save money. You bought a different kind of signup at a discount.

The narrower your product’s fit, the less comfort you should take from a broad pool of inexpensive trials. Let activation establish whether the extra reach helps before you scale it.

9. Give the ad set time to show you who activates

Meta needs roughly 50 optimisation events per ad set per week to exit learning, and edits that reset learning can start the process over again. Most SaaS budgets cannot buy 50 activations a week per ad set. Split $150 a day across ten ad sets and you give each one little room to learn, then make the problem worse by changing the budget every time a day looks ugly.

First step: consolidate to two or three ad sets and agree on a review point before launch. Give each ad set seven full days or 50 trial events before you start changing budgets, creative or audiences, unless something is plainly broken. Judge the comparison on cost per activated trial. An ad set at $28 per trial with three activations beats one at $12 with zero; you need to let both run long enough to see it.

There is a real trade-off here. If you cannot generate enough activation events for Meta to deliver against, use the earlier validated trial event for delivery while you build volume. But do not go back to treating cost per trial as success. Report activation by ad and audience, and keep looking for a qualified event you can send back reliably. That is a fallback for sparse data, not permission to buy tire-kickers forever.

Apply the same test when hiring someone. Ask which event they will optimise for and how they will report activated trials per $1,000 spent. If the pitch begins and ends with cheap signups, walk away. groas runs its Meta Ads service for a flat $1,499 a month rather than a cut of spend, with audiences, budgets and ads worked around the clock, actions logged and one human accountable for the outcome. The target is paid conversion, not a tidy trial dashboard.

If I could fix only one thing this week, I would start with No. 1: change what the campaign is trying to buy. The $15 trial set that produced 19 paid accounts from 140 trials beat the $9 set that produced four from 220. A higher cost per trial can be the receipt for finding real users. Pay it.