Your Meta Ads problem is measurement, not targeting
Before you rewrite your audiences, check whether Meta can actually see who converted. In most accounts we audit, it can't — and every optimisation downstream of that is guesswork.
There is a predictable conversation that happens when a Meta account stops performing. Someone suggests the audiences are stale. Someone else suggests the creative is fatigued. A third person suggests testing lookalikes at 3% instead of 1%. All three might be true. None of them is usually the actual problem.
Start with Event Match Quality
Open Events Manager and look at the Event Match Quality score on your primary conversion event. It's scored out of 10, and it tells you how well Meta can match the conversions you report to real people it knows. Below 6, the delivery system is optimising against a blurred picture. Below 4, it is close to guessing.
The score is driven by how many customer parameters you send with each event — hashed email, phone, name, city, IP, browser ID. A browser-only pixel typically sends two or three. A properly configured Conversions API implementation sends seven or eight. That difference is usually worth more than any audience change you could make.
Then check for double counting
Run a purchase or a lead on your own site and watch the event fire in the Test Events tool. If you have both a browser pixel and server-side events without a shared event_id, Meta counts the same conversion twice. Your reported ROAS looks better than reality, your true cost per acquisition is hidden, and the algorithm optimises toward a phantom.
Deduplication is a single shared identifier passed from both sources. It takes an afternoon to implement and it is the most common thing we find missing.
Only then, talk about creative
Once the signal is clean, creative genuinely is the lever — Meta's current delivery stack effectively derives the audience from the creative itself. But creative testing on top of broken measurement produces conclusions that don't survive contact with the next month.
The uncomfortable version of this advice: if your Event Match Quality is 4 and you're about to spend a month testing hooks, you will learn almost nothing. Spend that week on the measurement layer instead. It is boring, it doesn't look like marketing, and it is where the money is.
Written by
The Techlo Inno team
We build websites, paid media systems and CRM automation from Denver, Colorado. Everything above comes out of accounts we've audited or rebuilt.