Meta optimises for exactly what you ask. Most accounts ask for the wrong thing.

Meta's ad delivery system is extremely good at one thing: finding more of whatever you told it to find. That is its strength, and it is also why so many campaigns disappoint. The system did its job. The instruction was wrong.

Performance marketing on Meta has changed shape in the past few years. Targeting, which used to be where the skill lay, is now largely handled by the system. What is left to the advertiser comes down to three things: what you ask it to optimise for, the signal you send back, and the creative you give it to work with. The first two are where most accounts go wrong, and they are the cheapest to fix.

1. You asked for clicks and got clickers

Optimise for link clicks and Meta will find people who click. They are not the same people who buy, donate or book. If you want leads, optimise for leads. If you want purchases and have enough of them, optimise for purchases. An ad set chasing a cheaper, earlier action will happily report a low cost per click while your phone stays quiet.

2. You asked for leads and got anyone with a thumb

A lead form that asks for nothing gets filled in by people who want nothing. For a contractor, one added question, such as the type of job and the postcode, removes the enquiries that were never going to become work. Volume drops and cost per lead rises. Cost per booked job, the number that matters, usually falls.

3. The signal you send back is broken

The system learns from the conversions it can see. Two faults are common. Conversions are missing, because tracking runs in the browser only and some of it is blocked. Or they are counted twice, because the browser and the server both report the same event without a shared ID to pair them. Either way the system is learning from a wrong number, and so are you. Our tracking guide has a ten-minute check.

4. You split the budget until nothing could learn

The usual guidance is that an ad set needs around 50 results in a week to settle. Divide a small budget across five ad sets and none of them gets close. One ad set with the whole budget will usually beat five that are each starved. There is a budget calculator in our learning phase guide if you want to check your own numbers.

5. You are grading the result on the platform's own marking

Ads Manager credits itself generously, particularly for conversions by people who were merely shown an ad. The figure is useful for comparing one ad with another. It is a poor measure of what you earned. Check it against something Meta cannot influence: your CRM, your donation platform, your bank. If the platform says you had 80 conversions and your records say 50, plan around 50. More on this in our attribution guide.

The order to fix things in

1. Tracking. Until the count is right, nothing else can be judged.
2. The optimisation event. Ask for the outcome you want, or the closest one that happens often enough.
3. Structure. Fewer ad sets, each with enough budget to learn.
4. Creative. Now test. With the first three in place, a better ad shows up as a better result, and you can believe it.

Most accounts start at step four, because creative is the visible part. It is also the part whose results you cannot trust until the first three are sound. That is why we start every engagement with tracking.