Ads and traffic
Why ads burn the budget: 7 mistakes in Meta Ads
When a client comes to us saying “the ads don’t work”, in most cases the problem is not the algorithm and not that Meta broke down. The problem is specific decisions made at the start.
Below are the seven mistakes that occur in accounts most often, and what to do about each. All of them can be checked in one evening.
1. The campaign is restarted before it can learn
The most expensive mistake. The algorithm needs a certain number of target actions to leave the learning phase and deliver ads predictably. While it is in that phase, the cost per result jumps — and that is normal behaviour, not a fault.
Here is what an impatient advertiser does: sees an expensive lead on day three, switches the campaign off, creates a new one. Learning starts from zero. Three days later — the same thing. In the end the account spends months in permanent-start mode and never shows what it is capable of.
What to do: give the campaign at least a week without edits, unless the numbers are catastrophic. Make changes no more often than every few days, and one parameter at a time.
2. The budget is split too finely
A classic picture: five campaigns with five ad sets each, every one on a token daily budget. The logic is understandable — “let’s test everything at once”. The result is the opposite: each ad set gathers too little data, and none of them exits learning.
What to do: fewer ad sets, a bigger budget for each. Two ad sets that actually learned are better than twenty that never had the chance.
3. Optimising for the wrong event
The campaign is set to optimise for clicks or views, because there are many of them and they are cheap — the report looks good. But a click is not a business goal. The algorithm honestly does what it was told: it finds people inclined to click, not to buy. Those are different audiences.
What to do: optimise for the event closest to money that the account manages to accumulate in sufficient volume. If purchases are few — add to cart or lead, but not clicks.
4. The pixel is either missing or set up wrong
Without correctly configured events the algorithm works blind, and you see numbers that do not match reality. The most common breakdowns: the conversion event does not fire; it fires twice and duplicates leads; or it sits on a page a person reaches without submitting anything.
What to do: check the events before launch and again after any change on the site. Set up server-side event tracking separately, because part of the browser signals gets lost.
5. One creative for the whole campaign
Ads burn out. The same image shown for the fifth time stops working — the cost per result creeps up even though the settings are unchanged. That is not a broken account, that is audience fatigue.
What to do: keep several creatives in rotation and add new ones regularly. Look not only at cost per lead but also at frequency: when it rises while results get worse, it is time to refresh the material.
6. The ads lead to a page that is not ready to receive traffic
A campaign can be flawless and the money will still burn if the person lands on a slow page where it is unclear what to do next. This hits mobile traffic especially hard, and on social media that is the bulk of it.
What to do: check the loading speed from a phone, remove unnecessary fields from the form, make the main action visible without scrolling.
7. Success is measured by the wrong metric
The most insidious mistake, because it looks like working with data. The report shows reach, clicks, cost per click — everything is growing, everything looks fine. But there are no leads. Or there are leads, but nobody counts how many became clients.
What to do: define one main metric before launch — cost per lead or return on ad spend — and count it end to end, through to an actual payment. The rest of the metrics are for diagnostics, not for reporting.
What matters most here
Most “burned” budgets are not one fatal mistake but the sum of small ones. The campaign was not given time, the budget was scattered, the wrong event was optimised for, and then the wrong metric was used to judge it.
The good news is that all seven points can be checked quickly. The bad news is that after fixing them you have to wait again for the algorithm to learn. Patience here is as much a part of the work as the settings.
A case from practiceChildren’s clothing: 4.95 ROAS from adsView the case