Creative is the last lever with real headroom

Bidding and targeting have converged across every major platform; two competent buyers now get near-identical delivery. What still separates accounts is the concept, the hook, and how fast you can put a new one into the auction.

Ten years ago, a skilled media buyer could beat a competent one through account structure alone. Granular segmentation, manual bid adjustments, dayparting, placement exclusions — the work was real and the edge was durable.

That edge has largely been automated away. Every major platform now optimises delivery with machine learning that outperforms manual management in most conditions, and it does so identically for you and for whoever you compete with. Two competent buyers with the same budget and the same objective will land in roughly the same place.

What has not converged is what you put into the auction.

Creative is now the targeting

This is the shift that many accounts have not absorbed. On platforms with broad targeting and algorithmic delivery, the creative itself determines who sees the ad. An ad that opens with a specific problem will be delivered to people with that problem, because the early engagement signal teaches the system who responds.

Which means creative decisions are audience decisions. Choosing a hook is choosing a segment. Teams still running detailed interest-based targeting behind generic creative have the relationship exactly backwards.

Volume beats polish, up to a point

The single most reliable predictor of paid performance we see is not production value. It is the number of genuinely distinct concepts an account tests per month.

Note distinct. Five colourways of one concept is one concept. Five different arguments for why the product matters is five. Most accounts producing "twenty assets a month" are producing two or three ideas with variations, which is why the testing programme never surfaces anything new.

A creative testing cadence that works

Batch every two weeks. Three to five genuinely distinct concepts per batch, not variations.
Vary the hook, hold the offer. If both move, a winner tells you nothing about why.
Judge on hold rate and cost per incremental acquisition, not click-through rate.
Give each concept enough budget to exit the learning phase, or you are reading noise.
Retire on fatigue, not on age. A concept still producing is not "old", it is working.

The first three seconds carry the result

For video, hold rate in the opening seconds predicts final performance more reliably than any other early metric. If people leave immediately, nothing downstream can recover it — the platform will not distribute what nobody watches.

Hooks that consistently earn attention tend to do one of four things: state a specific problem the viewer recognises, contradict something they believe, show an outcome before explaining it, or open mid-action rather than with setup. Brand-first openings — logo, tagline, product beauty shot — perform worst almost universally, because they announce that an advertisement has begun.

Nobody grants an ad three seconds because it is well made. They grant it because the first frame implied the next one was worth waiting for.

Fatigue is a creative problem with a media symptom

Rising costs on a stable audience are usually read as a bidding issue. More often, frequency has climbed on a concept the audience has already absorbed.

The useful diagnostic is to watch cost per acquisition against frequency for each concept individually. When CPA climbs as frequency rises, that concept is spent. Raising bids at that point buys more impressions of something people have already decided to ignore.

The remedy is a queue. Accounts that maintain a backlog of tested concepts can replace a fatigued one immediately. Accounts that do not are forced to either keep spending into decline or cut budget while something new is produced — and the production lead time is where the quarter gets lost.

What this means for how teams are built

If creative is the constraint, the bottleneck is production throughput, not media expertise. That has organisational consequences most agencies are slow to admit:

Creative and media cannot sit in separate teams with a brief passed between them. The feedback loop is too slow to sustain a two-week cadence.
Production has to be built for volume and iteration, not for the flagship film. Different tooling, different budget shape, different definition of "finished".
The people making the ads need direct access to performance data. A creative team that learns what worked a month later is a creative team that cannot improve.

The uncomfortable implication is that a lot of paid media budget is currently paying for optimisation work the platforms already do. The headroom is upstream, in the ideas — and that is a harder thing to buy, which is precisely why it is still worth something.