Happy Monday! I am going to keep this one quick; I'm still 14 days our from surgery and not feeling great at the moment so don't have a lot in the tank. I know progress isn't linear and there will be lots of ups and downs, but the downs suck. Whenever I need to bang out a newsletter I like to do quick, random bullets with random things that are on my mind so here goes.
Before we do, many thanks to today's first sponsor Insense. I talk about ad creative and influencer seeding a lot, and for very good reason. They're both extremely important, and you need look no further than Insense. Insense is your all-in-one place to find, source, brief, and work with high quality content creators at scale for all of your ad creative needs. There's a reason Insense is trusted by over 1,500 DTC brands for their influencer and content needs; including fast growing brands like Solawave, GoPure, and WaveBlock. Insense is offering readers of my newsletter $200 off their first campaign when they book a demo right here.
PMAX isn't ready for primetime. There are two ways to look at PMAX. One is the very skeptical way of thinking that it's a black box that Google is using to sell us a bunch of junk inventory by combining it with the few bits of good inventory to give advertisers a minimum viable return. The less skeptical view is the idea that machine learning systems across a variety of different placements and devices while focused on an audience is the future. In theory, instead of trying to limit campaigns to placements, you can better convert users by creating audiences and showing ads wherever there are showing up across Google properties. I think a combination of theories makes the most sense right now, and I definitely urge every brand to run their own incrementality tests. But from what I've seen, there's just no good way to have PMAX be a true prospecting campaign and it's not going to be as incremental as a Youtube, Nonbrand Search, or Nonbrand shopping.
I still think too much ad variant testing is a waste of time When it comes to creative testing, I think there is not a ton of value in small iterations and learnings you get from them are invalid as there is a ton of machine learning noise. Even if you launch the same ad in the action twice, you are going to get drastically different results. If you really want to split test one variable in an ad to generate learnings, you need to run a proper in-platform AB test and spend enough on it for it to matter. That's also a really inefficient way to test and there will be wasted spend, so most marketers only do it if they really value learning. Any findings from variations launched in a different way won't be statistically valid so you should be really careful of generalizing findings based on this way of testing. If you want to know if you should use captions or not, don't launch the same ad with and without captions into the same ad set; run a proper AB test in Meta across a few different ads for consistency and run enough spend through it to get statistically significant data.
CRO is the opposite. You want to test in as structured way as possible to generate statistically significant learnings. You do this by only changing one variable at a time, and never more and running significant traffic through it. If you want to change more than one variable, you run a multivariate test and change just one of the variables between them, with coverage of all variables across the variants. Since there is no algorithm, you can be pretty confident in generalizing learnings from website and landing page tests. That makes it a great way to learn about your users and customers more. One thing I love to do is remove elements or sections to see if they matter or not. I think a lot of times were focused on adding conversion elements; but I love to remove features or sections to see if they even matter. Then you can decide whether you want to remove it to de-clutter, or add in something more important. For example, we had an ingredients section on a landing page we just removed with zero hit-to-conversion rate. Our heatmap data showed not a lot of interaction on ingredient sections so we tested.
Partner Of The Week - When it comes to website optimization, most marketers are still using the wrong data to make decisions. Legacy heatmaps, scrollmaps, and screen recordings are based on site visits and clicks. If you sell pageviews, that’s great. But we sell products. Heatmap is the only on-site analytics platform that tells you how to make more money. It shows you exactly how much revenue can be attributed to every element on every page of your website. This lets you measure the actions behind every transaction, optimize for revenue, then test and update and test some more.
Unreal Snacks, Obvi, Portland Leather, and more than 1,000 other brands are already using Heatmap to increase Revenue per Session. Right now, my readers can get 20% off a Pro plan using this link. Just use code CP20 when you sign up today.
Let me know your thoughts on some of these takes!
-Cody