Fundamentals

If your Linkedin or Twitter feeds look anything like mine, you’re painfully aware that, earlier this week, OpenAI released an update to ChatGPT’s 4o model.

And, if you went to try it on Friday, you found that the feature everyone was talking about was basically turned off because the company’s infrastructure couldn’t handle the demand. So, why the demand?

Well, image generation got really, really good.

My Twitter feed, in particular, featured growth marketers and agency owners creating hundreds of static meta ads, using existing ads as templates, slapping a quick prompt on it with a few of their own assets, and marveling at how good everything looked.

It was one of the coolest days I can remember for DTC Twitter in a long time. Such an incredible amount of experimentation and willingness to share. And while I love that part, and am equally marveling at how good these images are from a technical perspective, I am wondering a bit about how good these images are from a marketing perspective.

The macro thing here, to me, is that, yes, we’re going to produce a ton of ads using generative AI. Maybe we’ll even just use ChatGPT (hard to believe anything will be better any time soon, but it’s also hard to believe this even exists!). And while the idea of grabbing a template, and asking the model to adjust that original to fit your brand is incredibly exciting from a volume and velocity perspective, it likely means that we’re going to miss the mark in our advertising more than we do now.

And that’s because we’re rushing to explore what can be done as opposed to focusing on what should be done.

All this to say, the tools we’re using to do things are changing really fast, and that’s exciting. But, right now, in our race to explore, we’re leaving fundamentals behind. That’s OK for exploration, but it’s not OK for actual execution.

If our volume and velocity goes up, command of the fundamentals—knowing your customer, being about to understand the context in which they operate and are attracted to your product, brand and category—will either make the end result orders of magnitude better or worse.

To me, it seems like this is where consumer data really comes into play. There is the promise of personalization, yes, but the biggest value driver our industry has reached consensus on is not so much the personalized site experience; it’s the ability to hook a customer with an ad.

Those hooks, I’ve found, are always found in customer data: the surveys, the reviews, the purchasing habits that your customers share with you. Today, sorting through that is incredibly tough. That part is still, mostly, manual.

Maybe it won’t be manual for long, but right now it’s trailing ad generation. And if we don’t know our customers well enough to know which ad concepts to prompt, then we won’t be as impressed with the technology as we were earlier this week.