Split prospecting and retargeting

Prospecting advertises to people who do not know you. It is expensive per result, it fills the pool everything else draws from, and its return is realised later and elsewhere. Its job is supply.

Retargeting advertises to people who already came close. It is cheap per result, it has a hard ceiling set by how many such people exist, and a substantial fraction of what it is credited with would have happened without it. Its job is conversion rate on a fixed pool.

The two are not competing uses of the same money. One produces the input the other consumes, which is why treating them as interchangeable lines in an optimiser is how an account slowly stops growing while its reported numbers improve.

Three mechanisms, all structural.

Selection. The audience is people who already showed intent. Some proportion of them convert regardless, and the advertising is present at the moment they do.

Attribution. Last-touch reporting credits the final ad. Retargeting is almost always the final ad, so it is credited for conversions that prospecting generated and the retargeting merely witnessed.

Cheapness of the pool. A small audience, seen often, produces a low cost per result and a high frequency. Both look like efficiency and one of them is saturation.

None of that makes the numbers false. It makes them the wrong numbers for the decision being made. The question is not which line reports a better rate; it is what happens to the total if you move money between them, and reported rate does not answer that.

The one cheap check that does help: look at what the account's blended figures become with the retargeting line removed from them. If most of the apparent return sits in the smallest line by spend, the headline numbers are that line wearing a disguise.

The honest framing turns the split into arithmetic rather than preference.

Retargeting can only spend against an audience prospecting built. So the ceiling on retargeting spend is set by the rate at which the pool refills, and money above that ceiling buys frequency rather than reach — the same people, more often, at a rising cost per result.

That gives three numbers worth reading before dividing anything.

How fast the pool refills, which is prospecting reach over the period. Reach is reported de-duplicated, so one person seen on two days is one person across the week and two daily rows — anything adding daily reach is adding daily reach, which is a different quantity from the reach of the range, and the surface showing it says which.

What frequency is doing on the retargeting line. Rising frequency with flat results is the saturation signal, and it is readable from the counts without any threshold being asserted.

What the two cost you together, per result, blended. This is the number that actually moves when the split moves, and it is the only one that answers the question being asked.

Work the arithmetic on your own figures rather than a benchmark: the CPA calculator and the ROAS calculator take your margin and no account, and the break-even line they produce is what both lines have to clear together.

In a media plan this is two items, not one with a note. Each carries a platform, a name, a whole-flight budget in minor units and a flight inside the plan's window, and each becomes its own campaign draft with its own record.

Merging them produces an average that describes neither, and it produces it permanently: six weeks later the campaign's numbers are a blend nobody can unblend. Two items cost nothing extra and are the difference between a report you can read and one you cannot.

Name them for the job rather than the audience settings. Settings change; the job does not, and the item's name is carried straight onto the draft and then onto the campaign, so it is the name a report groups on.

The budget allocator template is the sheet version of the division, and How to write a media plan is the document both lines sit in.

Two rules about the mechanics, because they shape how the move is made.

A budget may not move more than a hundred per cent of its current value in one change. Upward that makes doubling the most a single change may do; downward a hundred per cent is the whole of it, so one change can take a line to zero. A reallocation is therefore at least two changes — one down, one up — and they are two rows on the record with two authorities, which is the correct shape for a decision you may want to explain later.

Applied increases on one connection are also summed over a rolling twenty-four hours against the account's ceiling, so a reallocation made as a dozen small nudges can be refused where the same money moved in two changes would not.

And a rule you wrote can clear the downward half on its own — approval rules clear pauses, budget decreases and bid decreases under a cap they state in a currency they state. The upward half always waits for a person. That asymmetry is deliberate: the reversible, spend-reducing direction is the half worth automating.

Which is the moment to state the arrangement plainly: your connections read your accounts, a change is written as a draft on the record with the state it would change from and the checks it passed, applying it is gated on approval — by you, or by a rule you wrote — and no platform write reaches a customer account until that platform has granted write access. Applying a cleared change to a platform is in private beta.

Nothing in reporting will, which is worth saying plainly rather than implying otherwise. The only evidence that answers "is this line incremental" is a test that withholds it from a comparable group and measures the difference, and running one is a decision about time and volume rather than about budget.

Short of that, the discipline is: size retargeting to the pool rather than to its reported rate, keep the two lines separate so the numbers stay readable, and grade the split on the blended cost per result rather than on either line's own. That is a worse answer than a real test and a much better one than following the flattering number.

Turn a media plan into campaign drafts is what happens once the split is decided, and Set a monthly budget across channels is the layer above it.