This looks pretty cool…creative testing is really difficult and sucks. It’s such a grind and this looks like it could fix that. I cofounded Agency Within (now WITHIN), the largest independent DMA in the US working w/ brands like Nike, Shake Shack, Goop + Rothy’s (so I kind of know what I’m talking about 🙂
We spent in the hundreds of millions of dollars on Facebook. A big part of why we were so successful is because of how data-driven / analytical we were. However, one of the areas that is really tough to bring that mindset of data analysis to is creative bc the data is so unstructured.
The current options are shitty:
+ Use FB’s dynamic creative tool (which imposes lots of other limitations)
+ Be really thorough and create multiple iterations of each creative with different backgrounds, text overlays, hero / product shots, landing pages, etc.
But these options suck: you are then left to FB to quickly decide the ‘winner’ and promote that ad. Very little useful data on what elements are outperforming, which audiences are responding to what messaging, etc. Basically you get ‘good results’ from the ‘test’ but are left without anywhere to turn in terms of what to test / iterate next. This is why even the best agencies throw their hands up and do the second version and then get stuck.
Some of the tools that I’ve researched, without naming any specific ones:
+ Claim to run your ads through ‘AI’ > but these are very basic rules bc they are high-level and not at all tied to your specific industry, audience, etc. > you end up with generic insights and suggestions
+ Help you templatize your ads > but the templates are driven by HiPPO (usually the Creative Director) and have not been iterated / tested at all > you end up with Mad Men for a modern age, not in a good way
+ Batch creative tools > these are useful in getting you more creative more efficiently > but then create a headache in that you have more to test without a structure for how to do that.
Any thoughts on how this compares? Just worried that I’m promised AI hands-off turn-key testing and get some bs outputs that aren’t that actionable or accurate.