Over the last two years we have audited more than 60 Meta ad accounts. The industries were different, budgets ranged from 20,000₺ to 900,000₺, some were run by an agency and some in-house.
Almost all of them turned up the same five mistakes.
What follows are not exotic technical details. None of them is a hidden setting. They are simply the things nobody looks at, because they are boring.
1. The pixel and the Conversions API count the same sale twice
This is the most expensive mistake on the list — because it corrupts every other decision you make.
Sales data reaches Meta through two channels: the pixel in the browser and the Conversions API from your server. Both report the same sale. If matching by order ID (event_id) has not been set up, Meta treats them as two separate sales.
The result: your dashboard shows 4.2x ROAS when the real figure is 2.1x.
Think about why that matters so much. You raise budget looking at the wrong number. You pick the “winning” creative looking at the wrong number. You scale a campaign that is losing money, looking at the wrong number.
How to check: Events Manager → your data source → the “Event Deduplication” section. If the match rate is low, or it says “no deduplication”, you have a problem. A rougher check: compare the number of sales in the Meta dashboard with the number in your e-commerce dashboard over the same date range. If the gap is more than 20%, look into it.
2. Frequency has passed 4 and nobody noticed
Frequency is the average number of times the same person sees your ad.
Up to 3 is fine. Between 4 and 5 is the red zone. At 6 and above you are paying money to annoy the same people.
Frequency climbs for one of two reasons: the audience is too narrow, or there is too little creative. The two usually arrive together — a narrow audience being served a handful of creatives for a long time.
The symptom is this: your CPM stays flat but CTR falls month after month, and the cost per sale quietly rises. You look at the chart and say “the market has tightened”. The market has not tightened; your audience is tired.
What to do: Widen the audience first — Meta’s algorithm goes blind on a narrow audience and works well on a broad one. Then increase the number of creatives. Those two on their own usually bring frequency down.
3. Too many campaigns, too many ad sets
One account we audited had 11 campaigns and 34 ad sets. The monthly budget was 162,000₺. That is an average of 160₺ per ad set per day.
To get out of the learning phase, Meta needs roughly 50 conversions per ad set per week. On 160₺ a day that is not possible. The result: none of the 34 ad sets ever leaves the learning phase, and all of them stay permanently in trial mode.
Splitting the budget feels like control. But on Meta, control comes from concentrating, not from splitting.
Roughly the right structure: Three campaigns is enough. Cold audience (about 70% of the budget), retargeting (about 20%), existing customers (about 10%). Keep the number of ad sets as low as you can.
4. The creative has been the same for three months
This is the most visible mistake and the most widely ignored.
On Meta, creative determines the bulk of performance. Targeting, bid strategy, campaign structure — all of them matter, but none of them as much as the creative.
A creative has a lifespan. Six to eight weeks on average. Then it tires: CTR drops, CPM rises, the cost per sale goes up. That is unavoidable. What is avoidable is carrying on with the same creative after it has tired.
A practical rule: At least three or four new variations should go live every week. Not a new concept from scratch — a different opening, a different first frame, a different headline on the winning creative. Once you have found the winner, the job is keeping it alive.
5. Writing a discount over a product photo
When a brand says “we produce creative”, this is usually what it means: a product photo on a white background with “30% OFF” written over it.
That is not an ad, it is a price tag.
Someone scrolling Instagram passes three or four pieces of content a second. Your job is to stop that scroll. A product photo on a white background does not stop anyone — because it looks exactly like an ad.
What does stop them: The moment the product is being used. A screenshot of a real customer review. A before-and-after comparison. A problem named in the first two seconds. A human face. Movement.
None of this needs a studio. Most of it is shot on a phone. Looking like it was shot on a phone is part of why it works.
Where to start
Do not try to fix all five at once. The order is:
Measurement first. Until the double counting is fixed, every improvement you make rests on guesswork. This is a day’s work.
Then structure. Simplify the campaigns and concentrate the budget. A week’s work.
Then creative. This one never ends. But this is where the gains are.
If you do not know which of these five are in your account, we run a free audit — we send the findings in writing, and the report is yours to keep even if you never work with us.