An ad outperforms. Finally! A winner.
Break out the champagne.
Everyone celebrates.
Why though?
The copywriter says that OBVIOUSLY it was the hook.
The owner says that it was the new offer which is totally "Crushing right now."
The media buyer is like, "Yeah right now video is doing really well."
It's all guesswork.
Because five things changed at once.
So the next batch of creative gets built on a guess wearing the costume of an insight.
Back when A/B testing was created and Hopkins wrote Scientific Advertising he was very adamant about changing ONE variable.
The man who wins out and survives does so only because of superior science and strategy.
Change one thing.
Isolate the variable.
Know exactly what you're testing before you spend a dollar to test it.
Win a test this way and you walk away with something you can use again.
Win it the sloppy way and you walk away with a good week, and nothing to show for it after that.
And all of that is well and good...
But being precise won't win you anything by itself.
The case for volume
Meta is rewarding volume right now. Feed the algorithm a wide spread of creative and it finds the winner faster than a small, careful batch ever will. A narrow test protects your ability to explain the result. A wide test protects your ability to find one worth explaining.
There's a second reason to go wide, and it has nothing to do with the algorithm. You don't actually know what's going to hook your market, not fully. You have a strong hypothesis, built from sales calls and gut feel and what's working for competitors. That's a good starting point. It's still just a guess until the market weighs in. I've watched an angle I was sure would flop run for four months straight, and watched the one I loved die on day two, more than once.
Test only your best guesses and you'll never find the ones you didn't think to guess.
What we're doing now
Build your creative in batches organized around a single angle. Inside that batch, change one variable at a time:
- Headlines
- Creative
- Body copy
Across batches, change the angle itself.
Run wide at the angle level. Run narrow at the execution level.
Name and tag everything before launch, not after. If a winner shows up three weeks in and you can't remember what was different about it, precision was never possible, no matter how careful you thought you were being.
That's the whole game.
Test wide enough to find a winner. Structure it tight enough to know why it won.
Most agencies pick one. We built our process around doing both. If your last "winning" test left you with a good week and no idea why, book a strategy call and we'll show you what that process looks like.