Retargeting vs Prospecting

Every advertising budget is divided between reaching people who have never heard of you and reaching people who have. The division is rarely chosen deliberately, and the default outcome is usually too much of the second.

The economics are not comparable

Retargeting reaches a small, warm, exhausted pool. Those people have already seen the offer, and the pool does not refill quickly; frequency rises, and each additional impression is worth progressively less. Prospecting reaches a large, cold pool where nobody has seen anything, and almost every impression there is worth something.

Because retargeting looks efficient on a per-click basis, it will absorb budget indefinitely if left alone. It is worth measuring it on cost per incremental result, which is a different and much less flattering number than cost per click.

The failure mode of too little retargeting

Attention that is never closed. A large prospecting audience sees the offer once and, in a category where consideration takes days, does not remember it. The measurement looks reasonable, because the platform credits some of those later conversions to the ad that started the sequence.

The failure mode of too much retargeting

Paying repeatedly for the same decision. When a retargeting audience is exhausted, frequency rises into the range where people find advertising irritating, and the marginal return on each additional impression goes to zero and then negative. The campaign continues to report a positive return because it is measuring its own attribution.

A workable starting split

Most of the budget on prospecting, with retargeting sized to a cap rather than a percentage. A cap is the important part: retargeting should have a defined ceiling based on how many people are actually available in the pool, so that the budget cannot expand into a segment that is being exhausted.

The test worth running

Reduce retargeting spend for a period and hold total spend steady. If total results barely move, the retargeting was not doing the work the platform claimed. If results fall sharply, the segment was genuinely load-bearing and the cap should be higher than it is. This is a small, cheap experiment that resolves a question no dashboard answers.