Why targeting matters less than it used to
Meta's Advantage+ campaigns and TikTok's automated targeting learn who responds to an ad from real behaviour: who watches, clicks and buys. Narrow interest stacks often just shrink the pool the algorithm can learn from and push up your costs. With a broad audience, the creative itself becomes the targeting: a skincare ad showing a gua sha routine quietly finds skincare buyers, a hook about tax time finds small-business owners.
A simple testing structure
- One campaign, broad audience. Keep location, age and exclusions sensible, and let the platform do the rest.
- Test concepts before variations. A concept is a different angle or reason to buy (a before/after, a problem-solution demo, a customer-style review, a “3 reasons why” listicle). Start with 3 to 5 concepts.
- Then iterate on the winner. Once a concept wins, make 3 to 5 variations of it: new hooks, new first frames, different on-screen text.
- Change one thing at a time, so you know what actually made the difference.
The first three seconds do most of the work
Most viewers decide within a couple of seconds whether to keep watching. Open on the most visual moment (the product in use, the transformation, the result) rather than a logo or a slow intro. Add bold on-screen text, since many people watch with the sound off.
How to read the results
- Hook rate (3-second views ÷ impressions): is the opening stopping the scroll?
- Hold rate and click-through rate: does the rest of the ad keep people interested?
- Cost per purchase or lead: the number that pays the bills.
Give each concept enough spend to judge fairly (as a rule of thumb, a couple of times your target cost per purchase) before switching it off. Judging too early is the most common way good ads get killed.
Keep the pipeline full
Even great ads wear out as the same people see them again. Plan to add fresh concepts and hooks every few weeks, so you're never scrambling when performance dips.