Creative is the new targeting.

For a decade, paid social was a targeting game. You won by finding the right lookalike, the right interest stack, the right exclusion list. That game is over — and the brands still playing it are paying for it in rising CPAs.

Across the accounts we manage at Kyro, the single biggest lever on performance over the past two years hasn’t been bidding, budget or audience structure. It’s been the ads themselves. When we shifted effort from audience tinkering to creative volume and variety, cost per acquisition fell by an average of 31% within a quarter. This article explains why that happens, and how to build a system that makes it repeatable.

Targeting didn’t die — it moved

Privacy changes, signal loss and the rise of fully automated campaign types have quietly removed most of the controls media buyers used to rely on. Broad audiences now routinely outperform narrow ones, because the platforms have more first-party data about their users than any interest list can describe.

But the algorithm still needs a starting point. It needs to know who this ad is for. With detailed targeting gone, the richest signal it has left is the creative: the words, faces, settings, pacing and problem the ad shows. In other words, targeting didn’t disappear. It moved inside the ad.

You can see this in any account that has been running for a while. Put the same broad audience behind two very different ads, say a founder talking straight to camera and a fast product demo, and the breakdown reports will show two different groups of buyers. Their ages, placements and even the time of day they convert won’t match. Nobody changed a setting. The creative did the targeting.

Your creative is now your audience definition. Every hook, face and frame tells the algorithm who to find.

How the algorithm reads your ad

Modern delivery systems analyse an ad before and after it launches. They look at the visual content and copy, then watch who stops, watches, clicks and converts. Within a few thousand impressions, they’ve built a picture of the kind of person who responds — and they go looking for more of them.

That has two big consequences for how you should work:

• Similar ads find similar people. Ten variations of the same concept with different button colours will reach the same pocket of your market and fatigue it fast.
• Different concepts open new audiences. An ad about saving time and an ad about looking good will be served to different people, even inside the same broad campaign.
• Weak creative can’t be rescued by settings. If the first three seconds don’t earn attention, the system learns that nobody cares and stops spending.

Building a creative testing system

Once you accept that creative is targeting, the job changes. You’re no longer making “an ad”; you’re running a pipeline that feeds the algorithm a steady supply of distinct ideas. Here is the structure we use.

1. Concepts before variations

A concept is a unique reason to buy, aimed at a specific person with a specific problem. A variation is a new hook, format or edit of the same concept. We plan in that order: three to five new concepts a month, each with three to four variations. That mix gives the algorithm real diversity without spreading budget too thin.

2. A brief that fits on one page

Every concept brief answers the same five questions, which keeps the whole team focused on the idea rather than the production:

1. Who is this for, in one sentence?
2. What are they struggling with right now?
3. What belief do we need to change?
4. What proof makes that change believable?
5. What should they feel in the first three seconds?

3. Test in a dedicated campaign

New concepts go into a separate testing campaign with a fixed daily budget, so they don’t compete with proven winners for spend. After seven days, or once each ad has had enough spend to judge fairly, winners graduate to the scaling campaign and losers are switched off. No exceptions and no favourites.

The metrics that actually matter

Judging creative on ROAS alone hides why an ad works or fails. We read results as a funnel, with each metric pointing at a specific part of the ad:

• Thumb-stop rate (3-second views ÷ impressions) tells you whether the hook works.
• Hold rate (15-second views ÷ 3-second views) tells you whether the story keeps people watching.
• Outbound CTR tells you whether the offer and call to action are clear.
• Cost per acquisition by concept tells you whether the idea sells, not just whether it entertains.

When a concept has a strong hold rate but a weak CTR, the idea is working and the offer isn’t — so we fix the end card, not the whole ad. That diagnosis is what turns a creative team from guessing into iterating.

We review these numbers every week in a shared dashboard, grouped by concept rather than by ad. Every result, good or bad, goes into a creative learnings log: what we tested, what happened and what we think it means. After six months, that log is worth more than any single winning ad, because it tells the next brief where to start.

Where to start on Monday

You don’t need a studio to put this into practice. Start by auditing your last 90 days of ads and grouping them by concept rather than by campaign. Most brands discover they’ve been running two or three ideas in dozens of outfits. Then brief three genuinely different concepts — a new person, a new problem, a new proof point — and give them a clean test.

The brands that win on paid social over the next few years won’t be the ones with the cleverest audience hacks. They’ll be the ones that treat creative as a system: planned, measured and fed into the algorithm every single week. The targeting is in the work now. Make it count.