Big and little ideas
When AI gives you essentially unlimited creative resource, you can do a lot of creative testing. But pretty much everyone still has a limited media budget, so how do you make the most of it?
We use a framework called "big and little ideas." It involves first testing big, propositional differences and only then testing little variational differences.
Now don't let the naming fool you - little variational differences can yield surprisingly big results. But they rarely give you an insight big enough to use as a springboard for the next wave of tests. They are little ideas. They have an impact, but they aren't scalable. They don't have a "so what". They die out after the test is over, so you have to start again at zero. This is why so many A/B testing projects produce amazing case studies but don't translate into amazing business outcomes. They are missing bigger ideas.
Here's how we use this framework when running tests with creative AI. They also apply even if you're using other creative approaches.
1. Aim to spend 30% of your budget on tests, the rest on scaling results. This takes discipline, and you might consider doing testing vs. scaling in different campaigns to manage it. The downside is that copying ads may cause you to lose social proof and "memory" associated with the ads and their performance on sophisticated channels like Facebook, so be aware of the tradeoff.
2. Start by mapping out and testing your big ideas. These are usually propositional - meaning they focus on either different potential audiences or different potential benefits. Creative AI can help generate these ideas. Sophisticated platforms like Facebook also use algorithms to automatically find the right people for the ideas, so the mechanics of practically targeting the right people for a given audience or benefit are no longer as much of a challenge as they once were. The main job is to come up with big, different ideas.
3. Once a big idea shows promise, work on testing little variations. This is where Creative AI can really shine - both because it produces endless variation without tiring and because predictive algorithms can spot and learn from small variations more granularly than humans can.
4. Combine fatigue management and testing into a routine. Constantly launching tests can be relentless and confusing. Distilling it down into a weekly or bi-weekly routine of (i) switching off losing or fatiguing ads and (ii) replacing them with new test ideas makes it easier. Creative AI can help by ensuring your new test ideas are always the "next best" based on what you've tested before.
