A winning campaign with three losing ads inside it

Short answer: a campaign number is an average, and averages hide their own contents. A campaign sitting at a healthy return is often one ad doing the work while three or four others quietly spend. Nothing on the campaign line tells you that, so the budget keeps flowing to all of them.

How the average deceives

Take a campaign that spent 20,000 and returned 60,000. Three times over. You would scale that.

Open it and you might find one ad that spent 6,000 and returned 44,000, and four ads that between them spent 14,000 and returned 16,000. The campaign is genuinely at three times. One ad is at over seven times, and the rest are barely above break-even before you count the cost of goods.

Scaling that campaign scales all five. You will spend more on the four that were never working, and the campaign average will fall, and it will look like scaling broke the campaign. It did not. It just diluted the one ad that was carrying it.

Three places the loss usually hides

• Creative. The most common by a distance. One image or one hook does the work. The others were made in the same batch, launched on the same day, and never separately judged.
• Audience. A broad audience and a tight one inside the same campaign will rarely perform alike, and the campaign average sits between them, describing neither.
• Placement. The same ad in a feed and in a full-screen placement is two different ads in practice. One of them is usually much cheaper per order, and the split is invisible until you look.

The pattern is the same in all three: something inside the campaign varies a lot, and the number you are shown is the middle of it.

Why platforms do not solve this for you

They partly do, and it is worth being fair about it. Automated delivery does shift budget toward what performs, and it is usually better at that than a person checking weekly.

But it optimises toward the event it was told to count. If your purchase events are arriving without a value, or are being double counted, or are missing entirely for part of your traffic, then it is optimising confidently toward a distorted picture. The inside of the campaign gets reorganised around a number that was wrong to begin with.

That is the real reason to look inside yourself: not because you can beat the algorithm at allocation, but because you are the only one who can tell whether the feedback it is learning from is true.

Where MIQAS comes in

Campaign Lens opens any campaign into a two-week breakdown, so the question moves from “is this campaign working” to “which part of it is”. Performance leaderboards do the same for products and pages, which is where the other half of the answer usually sits.

The important difference is what the numbers are built from. Revenue comes from your actual orders rather than from what each platform claims, so an ad that looks strong only because its conversions are being counted twice does not survive the comparison.

Both are available from Starter, because this is not an advanced question.

What to do this week

• Take your largest campaign by spend. List every ad inside it with its spend and its orders. Not its clicks.
• Find the ad with the highest spend and the lowest orders. That single line is usually the whole story.
• Before pausing it, check it is not a measurement fault: an ad sending traffic to a page where tracking is broken looks identical to an ad nobody responds to.
• Then pause one thing, not four, and give it long enough to read. Changing everything at once means learning nothing.

A campaign average is a summary, and summaries are for reporting. Decisions belong one level down.