Austin Hay: observation

Transcript context

…them the ability to counter a conversion. Every ad network out there has little things they stuff into your URL that tell you that you came from them. Facebook has FVP, FPID, they sometimes encode it. Google has this thing called Google Click ID, which is just a really long string of characters that don't matter unless you know how to decode it. But all advertisers, and for the longest time advertising worked by putting parameters in URLs, pushing somebody through to your website, collecting those parameters and then passing it back to the ad network so they could get credit for it. And so in my mind, the best practice that everybody should stand up from day one is to basically design the system for MTA and then use whatever makes sense as you grow. And so the way that I typically recommend to people is like imagine when a user comes to your website, you collect the URL, collect the referring URL, collect all the additional marketing parameters that you might want, [inaudible 00:45:57], TikTok ID, Microsoft ID, you should just make a list of them. And if you don't have that list, I can give them to you. And then you should collect all UTMs. So, in the URL, you're going to have UTM campaign, UTM medium. Most marketers use this to note what the campaign type was. Now the thing is that UTM is only going to be specific to the moment in time that the person came to your website. So, back to example about Lenny's podcast. If I come to LE's podcast and I came from a Google ad, then my UTM is only for that Google ad. So, I have a Google Click ID and I have a UTM. So, what you're got to do is you've got to store those parameters locally on the device. Either was a browser or whatever. You got to store it as UTM first campaign, UTM last campaign. And what you do is every single time that a person comes, you replace the last campaign or the last value with the one that's there. So, say the last one was Facebook and then I come later from a direct mailer ad you replace the UTM last medium with the new one. Now what's happening if you're using third party tools is that you're collecting this user information when the person's on the website, you're going to collect it both as a user attribute and as an event that way. What's going to happen on the backend for your data warehouse team is they're going to see a user profile that has both the first attribution information and the last attribution information. And then for all the stuff in the middle, you're firing off a page view event with first and last, where the last might deviate if there were multiple steps in the middle. So, what they can do is they can just coalesce over all the last UTMs they've seen on all your events by user to get both their first one, all the ones in the middle and the last. And so this isn't actually that complicated to set up. Most people just don't do the work early on. And then when they want to go back later and have MTA results, they don't have the data to do it. So, one of the things I tell people who are debating this is let's just get the infrastructure right from the beginning. want to go back later and have MTA results, they don't have the data to do it. So, one of the things I tell people who are debating this is let's just get the infrastructure right from the beginning. Let's set up so that you have users, you have user attributes, you're collecting first and last UTM on users. You're firing events with all those. There's some other more complex things you can do too. You can set them in first party cookies and you can also set them in your third party cookies for your tooling vendors. At the end of the day though, what matters is you just are collecting this information from the beginning. That way when you actually want to progress your attribution model, you don't have to wait a really long time to start gathering that data. Amazing. I love the details that you're sharing. I don't know where else people can find this sort of advice. It sounds like a core part of this is one, just having a data warehouse where you just throw all this data into, and two, having a taxonomy that you can rely on and do multiple things with down the road. Is that roughly right?…

Stored transcript either side of the excerpt. The highlighted words are the published quote; the surrounding text is unedited source, never generated.