Target ROAS: Setting A Number The System Can Actually Hit

The short version

Target ROAS tells the bidding system what return you want per unit of spend, and it bids toward impressions predicted to deliver that ratio.

The part people misread: it is a constraint, not an aspiration. A target you cannot support means the system finds few impressions it predicts will clear it, so it serves less — or nothing. You do not get the same volume at a better return. You get less volume.

How to set it

Work from what you can afford, not from what you want.

Establish break-even ROAS from your margin. Below this you lose money.
Add the margin you actually need for the business to work.
Compare against what the account currently achieves.

If your target is far above current performance, you are not setting a goal — you are switching delivery off gradually. A realistic starting point is at or slightly above what the campaign already delivers, then tightened in steps.

The volume trade

This is the whole mechanic and it is worth stating plainly:

Target	What happens
Below current performance	More volume, lower efficiency
At current performance	Roughly current volume, stabilized
Slightly above	Less volume, better efficiency
Far above	Delivery collapses

There is no setting that produces more volume and better efficiency. Every move along this scale trades one for the other, and choosing which you need is the actual decision.

Moving it without breaking things

Change in small steps. Large changes reset learning, and a campaign that keeps re-learning never stabilizes.
Wait between changes. Give it enough conversion volume to respond before judging.
Change one thing. Adjusting target and budget together makes the result unreadable.

When target ROAS is the wrong tool

Insufficient conversion volume. The system needs enough events to predict value. Below that, targets are noise.
Highly variable order values. If the value distribution is wide, average-based targeting is a poor fit.
Long delays between click and revenue. If revenue arrives outside the window, the system optimizes on a fraction of the truth.
New campaigns with no history. Let it gather data before constraining it.

What it will not do

It will not make unprofitable inventory profitable. It allocates toward impressions predicted to hit your ratio; if few exist at your price, the answer is fewer impressions, not better ones.

And it optimizes toward reported revenue. If your tracking under-reports, the system optimizes toward an understated picture and behaves more conservatively than it should.

Practical rule

Set the target from break-even plus required margin, sanity-check it against what the account already achieves, and move it in small steps. A target far above current performance is a delivery switch, not a goal.