Log inRequest access

Placement concentration

Warningplacement_concentration5 min read

In shortAutomatic placements let delivery buy the cheapest surface, so budget quietly concentrates on one of them — and because the ad set reports a single blended cost per result, the placement that took the money and the placement that produced the results are never compared.

Leave placements on automatic and you have not chosen where your ads run. You have delegated it to a system optimising for the cheapest way to spend your budget, which is not the same as the best way.

That is often fine. It stops being fine when it becomes total: one live campaign put ninety percent of its spend into Instagram Reels, out of four placements the ad set allowed. Nobody decided that. Reels was simply the cheapest surface to fill.

Why the report will not raise it

Nothing has gone wrong from Meta’s side, so there is no alert. The campaign is delivering, within its settings, at the price the auction set. The ad set reports one cost per result, and that number is the average of every surface weighted by how much each one spent.

Which means a placement holding three quarters of the budget essentially is the number. If it performs badly, the ad set looks bad and the other placements are blamed for it. If it performs well, the ad set looks fine and nobody asks whether a surface with two percent of spend was quietly better.

A blended average across placements with a 90/10 split is not an average. It is one placement’s number with a rounding error attached.

Split it, do not exclude it

The instinct on seeing concentration is to exclude the greedy placement. That is usually the wrong move, and it is wrong for an informative reason: you do not yet know whether the surface was bad or merely cheap.

Splitting the ad set by placement gives each one its own budget and its own number. Two weeks later you have evidence instead of a theory, and the exclusion — if it is still warranted — is a decision rather than a reflex.

It does cost something. Each new ad set re-enters the learning phase, and splitting one into four means paying that four times over smaller budgets. Worth it when the concentration is severe and the spend is material; not worth it on a campaign running two thousand rupees a week.

What to check

  • Read the placement breakdown before the creative report. A poor blended result may belong to one surface.
  • Compare share of spend against share of results. A placement taking 75% of budget and producing 30% of results is the whole story.
  • Look for the Audience Network specifically. It is the cheapest surface and the one most often taking budget nobody assigned it.
  • If you split, split once and wait. Splitting, judging early and re-merging pays the learning cost twice for no information.

When Admetriq raises it

The actual rule, so you can hold this page against what you see rather than take its word for it.

  • Fires when the top placement takes 75% or more of an ad set’s spend (CONCENTRATION_SHARE = 0.75).
  • The finding names the surface, the share, and how many placements the ad set actually allows, so the sentence reads as what it is: this much of the budget went to one of these several options.
  • Set below the case that motivated it. One live campaign put 90% of its spend into Instagram Reels out of four allowed placements — not chosen, just cheapest to fill.
  • The proposed fix is to split the ad set rather than to exclude the placement, because splitting is what makes each surface measurable instead of guessed at.

Most often mistaken for

Low click-through ratelow_ctr

A blended click rate can be dragged down by one surface while the others are fine. If most of the spend sits in a single placement, a poor click rate is that placement’s click rate and the creative may be blameless — so check the concentration before rewriting anything, or you will replace an ad that was working everywhere it was given a chance.

All findings · Longer writing