So What? UTM Governance in GA4

Well, hi there everyone. Happy Thursday. Welcome to So What Marketing Analytics and Insights Live Show. I'm Katie, joined by Chris and John, and apparently I'm joined by my dog this week, uh, who is behind me with itchy ears. Uh this week we are talking about UTM governance in Google Analytics 4.

We've been trying to cover a lot of different aspects of Google Analytics 4 because there's a lot to it. It's essentially a whole different system from Universal Analytics. And so we're trying to be as thorough as possible. And one of the things that is consistent, but also different in the new system is UTM governance. And so UTM codes uh are gonna be your best bet for making sure your data is coming in clean.

That's true in universal analytics. That's still true in Google Analytics 4, but we're gonna get into why it's a bit different this time around. So Chris, where would you like to start? I think we should probably start this week talking about just even how does Google Analytics make decisions about attribution? Because if we don't understand that, then it's difficult to understand like why UTM governance matters so much.

So we're gonna go ahead uh and share this flow chart that we made. Um and this is available in our Slack group. If you go to uh trustInsights.ai/slash analytics for marketers, you can join the free Slack group there. Uh this is just a one-page PDF. And here's what happens as uh data comes into Google Analytics.

The first thing that happens is uh an event occurs, right? And Google looks at it and says, Is there a g clip? A g clip is Google's uh uh click ID. If that's there, Google says, Hey, I know what this is. This is a Google ad.

So it's going to assign a source of Google ads, a medium of CBC, uh, any other UTMs that might be present, but at that point, Google's attribution efforts stop. They said, got it, we're all set. Unsurprisingly, Google gives strong preference to its own advertising system first. It's weird how that works. Secure the bag of money, right?

That's part one. Exactly. Second thing, if there is no G Clip, Google Analytics 4 says, okay, are there campaign tags, right? So there are certain campaign uh parameters that other advertising systems use. Is there a campaign source, a campaign medium, and etc.?

If there is, Google says, okay, that's an advertising system. Let's let's give attribution to that. You will see if you are in like Facebook ads or in uh other uh ad platforms, the ability to set a campaign source campaign medium. Um, these are different than UTMs, but they're structurally essentially the same thing. So Google looks for other paid ads second.

Again, makes total sense because part of the reason people focus so heavily on attribution is hey, we're spending a lot of money. Kind of want to know what's going on. Third, um, if there are UTM codes, Google looks at them and says, Are they properly formatted? If they are not properly formatted, you fall all the way down. This is like shoots and ladders as a kid, right?

Down into the ghetto. You go into sources, direct medium is none, other UTMs are not set, and you just cry a lot because you've lost your attribution data. So a properly formatted UTM code means that you're using things like you know the ampersands between the codes. They're not the names are not misspelled, like UTM source, but you mean you abbreviate it SRC. Google doesn't know what to do with those things.

So it's got to be the proper format. If that's the case, that's the next thing Google says. Okay, if it's all good, I'll assign it with the what it what you've provided. So we say a lot, I say a lot. Um UTM sort of a mnemonic abbreviation.

You tell me, Google's asking us, you tell me what to do with this data. So if you've got those UTM codes, Google will assign them and all will be good. That's where this is sort of like the the Google knows what to do bucket, right? Sort of the campaign stage. Once you're out of that, like you didn't include it, if there's no gclit, if there's no campaign tags, if there's no UTM tags, then Google goes into what we call the guessing stage.

And this is where things get ugly. If Google can detect a domain that the traffic is coming from, uh, it will then go through a process of looking at a domain and saying, Do I know this domain? And if uh Google has provided a long list of domains. So if there's no UTM codes and Google can see a domain, it will then it will essentially guess. It will say, like, hey, something came in from academia.edu.

I'm gonna categorize that as social. So in your attribution will show up as uh a source medium of essentially of social. Uh if we scroll down here, if if traffic is coming from, let's take a look here, uh duckDuckGo, right? It'll say, Hey, that's a search engine. So I'm gonna sign that uh category of search.

Uh if it comes from Facebook, uh, I'm gonna assume that it's it's social. And so at this point, Google is now guessing, trying to categorize its traffic. So if it makes a match, it will assign its own source medium from that list. That is downloadable from Google, that spreadsheet, and they update it reasonably frequently. If there's no known domain, Google says the source is the referring URL, the medium is referral, and then if there's UTM codes, they'll it'll attach them, but uh at this point there there aren't.

And if Google can't tell what the referring source is, it will look to see if it recognizes the visitor, the user. If there is a user, it checks the campaign timeout, which is a setting you set in Google Analytics. And if there is, if it's before the campaign timeout period, whatever the previous source medium was for that user gets assigned again. Google is basically saying, hey, we don't know how this person got here today, but we've seen them before. So we're just gonna give credit to the last thing we know they came from, if we have none of this other information.

And if either one of those cases is no, you're back to you know a direct uh direct and nut. Basically, Google says, I give up, I have no idea where this where this traffic is from, and this attribution is it just goes into nothing. So that's how Google Analytics 4 makes decisions about attribution. And the reason that this is important is because what you see is if you provide good UTM tracking codes, Google doesn't have to guess. Right?

And that's that's critical. If you provide good UTM tracking codes, Google does not have to guess. Now there's a second layer to this. If you're in Google Analytics 4, uh you will note, uh, in fact, let me bring up a a tab here of Google Analytics 4. You'll note that in the attribution section, Google Analytics 4 will say, I know I'm going to give you the opportunity to look at some channel groupings.

So let's pull up Google Analytics 4. Go into the advertising section, go into conversion paths. And here there's these default channel groupings. Now, if you've used Google Analytics 3, aka Universal Analytics in the past, you know that you could customize your channel groupings to make sure certain things with certain UTM codes fell into certain buckets. You can't do that anymore.

And they've published those rules in this really long post that says, you know, here are the conditions for what for how channel groupings are decided. And you'll see things in here, like for example, uh organic social, that long list of URLs. Uh it's going to guess, or if a medium is one of these things. Remember that that red when it says regex of social sites, that's back in that guessing stage in the flow chart. If we provide a good UTM, you good UTM parameters like Facebook and social, Google will put our traffic in the correct bucket.

And then when we go into Google Analytics for our attribution tool, the channel groupings will be correct. So that's why this governance is so important because you can't customize channel groupings anymore. You have to use the ones that are built in based on Google's rules. Is that so in universal analytics, everything had to be lowercase in uh the UTM tags? Is that true in Google Analytics 4?

Now says uh UTM uh uh case that channel definitions are are not case sensitive and cannot be edited. The channel definitions can't. I'm talking about what you put into your UTM tag. Our our assumption though is that based on this, the the conditions for these channel matchings is not case sensitive, uh which would make sense. Google's general regular expressions are not case sensitive.

So that was only sort of an anomaly in universal analytics. Again, because yeah, uh universal analytics was actually not built by Google. So this is with GA4 is essentially Firebase. It was built by Google. But then so is it that if you end up do using uppercase, it will default and treat everything as lowercase, or if it's uppercase, will it get tossed?

It just treats everything as lowercase, is our understanding because it says it's not case sensitive. That said, the best practice just for your own sanity is to make everything lowercase. Yeah, to go with that. How about as far as the inbound domains? Isn't that the kind of stuff that's getting blocked at the browser level just about everywhere?

I mean, is that eventually gonna dry up, or is there some other method they're using to find out the where those domain traffic, you know, who the referrals are? Referring string is usually brought in at the browser level. Um, so unless you have a specific ad blocker that's blocking that, um, that data should still be available. The exception is if you are going from a secure site, we just using SSL to a non-secure site, um, you will lose that data. But the if the digital certificate, if the SSL connection remains intact, then the referring traffic comes along with the referring domain comes along with it.

Oh, yeah, and that's a good point because it's yeah, ad blockers take care of that, but we're talking about the whole web. So there can be thousands of legit traffic that can go through. Exactly. So that's why this stuff is so important because you can't customize channel groupings. So in your attribution models, if you're doing channel level attribution, you've got to get your UTM codes correct.

So one of the uh pieces of advice that we would give clients, uh, especially around governance, is in order to preserve your UTM codes, the way that you have them set is to use a link shortener. And that does two things. One is it makes sure that certain sites don't strip out the UTM codes, but also so that extra data of UTM codes don't get added on to confuse Google. Is that still the same piece of advice that we would want to give when using Google Analytics for? Yes, absolutely.

The last UTM code that's um received is the one that's treated as authoritative. So like if you say a bit.ly link and and a site or a service attaches UTM codes to it, once the bit.ly link is expanded and your own UTM codes are embedded, those ones have precedence because they're the last thing that Google see that, you know, the last thing that it gets transmitted. So any previously added on crap gets tossed away and your valid ones get interpreted. So you it's a great idea to use URL shorteners. The other thing is a great idea to do is to have some kind of um process for putting these things together that doesn't require somebody hand typing them all the time.

So this is a stupid simple example of just a spreadsheet. You're putting your landing page URL, you choose your source, which we always recommend be the domain name, uh, and then uh you have you build yourself, and you can do this in Excel, you could do this in Google Sheets, it doesn't matter. You build a list of your known mediums, because we know from what Google is saying here, this is how medium is predominantly how Google makes decisions about non-paid stuff. So if you have uh correct mediums pre-programmed into a sheet, and then you and your team and your agency are all using the sheet, the chances of sneaking in something that doesn't belong there is much lower. Particularly if you don't let people type things in, they have to choose from a drop-down.

And I think that that's definitely one of the things that we see go sideways is when you have not only your team but other external agencies working on your behalf, that governance becomes even more important because they may have their own way of doing things, or you know, newer team members may say, Well, at my previous job, I did it this way. You take the guessing work out of it and say, this is how we at Trust Insights, this is our governance for UTM codes, this is how you have to do it. There's full transparency. You can see the codes that are getting built, you can see if they're not being put up correctly. Uh, and that's those are solvable problems that you can then fix with those team members or those agencies uh to say you're kind of messing with our attribution if you're not doing it the way that we've asked you to do.

So we need you to fix that. Exactly. Uh, you might notice that I just hit delete on two columns on the spreadsheet. Um, and you may be wondering like, well, why would you do that? That seems like uh a bad practice.

Well, the reason for that is that uh the content and keyword parameters, they're still interpreted on the uh on the back end, but if you look in Google Analytics itself, when you look at your things like uh choices you can make, go away, help box and ask you. You can see default oh go away, that's so annoying. There's uh default channel grouping source medium or campaign. So content and keyword, which are previously available as UTM tracking codes uh for analysis are gone. So if you're trying to dissect you know specific types of uh ad copy and stuff like that, you have to incorporate it probably in at the in the campaign UTM code now.

Uh put it all in campaign instead of having content and keyword. And so if you include content and keyword in your UTM string, is there nowhere to see that at all anymore in Google Analytics 4? Uh because I know a lot of agencies, especially ad agencies, use those strings to make sure that they know you know which ads are which. And so this is, I mean, much like everything with Google Analytics 4, it's change management. You're doing things completely different now.

Let's see. So if you're comfortable with Explore Hub, yes. If you're not comfortable with Explore Hub, then no. Is it are those metrics that you could look at in Data Studio, are they available there? Let's go find out.

Because it's it's it occurs to me that you know a lot of companies, teams, agencies, you know, you name it, have really tried hard to get really good UTM governance. And now Google's saying, great, now that you've mastered that, forget everything you know and do it a different way. And so I'm just sort of wondering if there is some sort of a happy medium so that people don't have to completely overhaul everything that they've been trying to teach their team to do. And if the content type is available, but keyword is not. Interesting.

Well, it's not available as of today. As of today, although I did just hit refresh. So as right now is not there. Yeah. Which isn't to say it won't be, but we Google has not let us know anything, not us specifically, everyone specifically, uh what they will and won't be including.

Exactly. So our guidance there would be um if you have what if you're used to in content and keyword, incorporate that into your campaign. And then when you're doing reporting, like in Data Studio, um, you can you can create regex filters to pull apart your combined uh you know campaignslash keyword thing into into something uh more easily interpreted. Not as straightforward as it was. So that covers you know uh UTM governance, so the basic of it and why it's so important because it is a lot more complicated than used it used to be.

The other thing I know we wanted to talk about is keeping an eye on this stuff to to understand like what's going on. Um sort of a little bit of governance. So in Data Studio, one of the things that we can do is we can actually build an understanding of our our different uh UTMs. So let's go ahead and do this. I'm gonna just make three tables here.

And what we're doing, what we're gonna be building here is just looking at are we using intelligent, sensible UTM codes. Remember, there's three different sets of source medium campaign. There is the first touch, which is called first user acquis uh uh acquisition. So first user source, first user uh medium, first user campaign. Then there is session level, which is what's happening in intermediate visits, and then there is what's called um conversion source medium and campaign.

So, what was the source medium and campaign at the time of conversion? So let's use we're using the trust insights account here. Let's do source. Yes, I'm in G A for our first user uh touch here is called user source, right? Uh let's use a source.

Uh let's do user medium and user campaign. So this is the first touch uh source medium and campaign. Uh let's go ahead and resize the columns to fit to data so it's a little bit easier to see. And so this is essentially for our website the first touch that people are coming in on. And we can see there's uh there's a decent amount of direct traffic, Google organic, uh, and then lots of different campaigns.

So that's uh and what we're looking for here are things that don't make sense that are are are messed up. Um scrolling down here. Like here's a couple. Like almost timely is a uh LinkedIn is a source, but there's no medium set. Right.

So that tells me that I did something wrong in my own uh posting of my personal newsletter on LinkedIn. I must have hosed the tags and and it didn't come in correctly. So from a governance perspective, that's wrong. Uh same here for uh my personal social media. I've got two not sets.

Again, that's something that should not be the case. Uh so I would need to go fix that. So that's our first touch. Let's look at our session level data. Now let's do session source, session medium, session campaign.

Get rid of these other two. And you'll notice the numbers here are different, right? Um, the the uh order of things are different because again, this is what's happening sort of in the middle, uh, essentially of it's like a mini customer journey almost. Um fit to data, and then for our last table, we're gonna use the uh source. This is confusingly not labeled conversion source medium.

Um it's just labeled because I don't know why Google did not label it properly. Uh oh, this is hard to see, so let's do a bit of theme layout and just make this a little bit wider. That's too wide. And we're going to resize colon's fit to data. Okay.

Now you'll notice um, this is a whole bunch of zeros, right? I think, because I don't think I have it set to um I don't think it have it set. Oh, this is views, right? The metric is not used. Metric should be events.

Or yeah. Okay. So this is essentially, let's go to view mode here. We're looking for anomalies, we're looking for bugs, right? Looking for things that that don't make sense.

Uh and if they don't make sense, we should try to figure out why they don't make sense. Um again, some of those things like the CSP newsletter not being set properly, that's uh a screw up, and that's something that we can go fix. So that's how you do your some basic uh analysis of your sources and mediums and all of your UTM codes, all three of your UTM codes. This will tell you what's what's broken and what's not. Well, and that means that you have to know what you're looking for, which goes back to getting familiar with the new world order of how Google is categorizing everything that's coming into your website, which is it's similar but just different enough from universal analytics.

So all of your standard things like Facebook, Twitter, LinkedIn, those are still social media. You know, email is still email, you know, and so on and so forth, but there's gonna be other gotchas. There I think there's new categories now uh of sources. Um of both, actually, you know, and you know, new social platforms get added all the time. So it's definitely something that you have to try to stay on top of.

Um, which is not easy. That in and of itself could be, you know, a decent amount of time of just staying up to date on what Google is saying. This is what's valid and this is what isn't. There's some obvious things like not set and direct or none. Uh, those are pretty straightforward to be able to pick them out.

Exactly. Um, and and that's really what you want to go after first. You want to hunt hunt down the easy stuff first, um, clean that up. Uh, another one that is important to look at is just that direct none bucket. Like just how big is that bucket?

And do you know why? There are some businesses and industries where that's gonna be the case. For example, if you are a B2B company in say cybersecurity, your attribution data is gonna look like almost all direct none. Why? Because all your customers are paranoid privacy nuts who have like 18 ad blockers turned on uh and you know are using super lockdown browsers.

So you're gonna get no attribution data. And there's no way to fix that. Like you're not gonna you're not gonna fix that. The only way you're gonna get any data from those folks is at you know, at the time when someone fills out a form. You say, okay, please fill out this form so we can get in touch with you.

Um and even then, it you know, even money, whether they lie or not. But um that's you know, that's gonna be the case uh in other cases like finance or healthcare, where people are using browsers and computers that are um behind a corporate firewall that's heavily screening stuff. Again, stuff's gonna get blocked. Um you may be doing business in certain locations like China, where uh some information is prohibited to be transferred, and so you may get you know less and less information from those places. So, in general, uh you can expect that you're gonna get less information.

Um while we used to say that direct traffic was a direct result of not doing attribution, you know, not doing governance well, that's not always the case anymore. So let's say I'm looking at this. Let's say, you know, I'm not at Trust Insights where we have a lot of control over this data and we can fix things. Let's say I work at a larger enterprise size company, and you know, a lot of this data is coming through incorrectly. Um, you know, I can go hunt down the individual people, agencies who are doing it incorrectly and fix it, but it's only correct moving forward.

I can't do anything about the historical data in Data Studio. Is that correct? Yeah, that's always the case. You can never go back in time. So how do you I mean, I guess, and that's always been sort of the the so what, you know, it's like, well, what do you do with all of that incorrect data?

It is what it is, you know, and so I guess I don't even know what I'm trying to ask because it's such it's like it's it's weighing down my brain in such a way of you put an analysis like this together, you see that everything is being coded incorrectly, you go ahead and fix it, but then you can't compare the fixed data to the older data, which may have been broken for a long time. It depends on the decision you're trying to make. Like if you're if you're just poking around, then it's not a big deal, right? It's a but if there's a decision you're trying to make, I had I I had a call this morning um with a prospective customer, and they're like, yeah, our data's all hosed. I said, you know, what decision are you trying to make?

And and the person on the phone said, uh, well, we're not really making any decisions. I said, okay, then you don't need to worry about your broken data because you're not using it for anything. It's like, I'm not sure if I'm not sure if this bread is bad or not. Yeah, but you're gonna eat a sandwich? No.

Okay, well, then I you know, it doesn't really matter if the bread's bad or not, uh, if you're not gonna use if you're not ever gonna eat it. Um if you're trying to say a very common use case is you're trying to make budget decisions about uh you know what to do, and all of your data is busted. Um you can do if you've got time, even just seven days worth of data, uh, once you've fixed everything is just do a uh a comparison. What is the you know what is seven days versus you know uh uh the previous seven days before you fixed everything look like? Are there sources of mediums that you don't have to fix?

So for example, um organic search traffic is not something you you can't fix it, right? You there's it's literally not under your control. Referral traffic from other sites is not under your control. If you can get a sense of how clean those things are, if they are consistent, then you know that the past data, you might be able to rearrange it. Like if you export it and you're processing it in an offline system, you can you might be able to fix it up to be apples to apples with the new the the new uh world order.

If on the other hand, maybe it was something as severe as, oh, we forgot to put tag manager on our website, then there's no fixing that. And you just kind of start making decisions with the data that you have. So and John, I would like to get your opinion on this as well. But you know, so Chris, you're giving the example of, well, you know, you're not making any decisions with the data, so it doesn't matter if your data's broken. I disagree with you because just because you're not making decisions with it today doesn't mean that tomorrow something isn't going to change.

And so I do think it's important to, you know, every if you're collecting the data, why would you be collecting it wrong? Period. You know, you either collect it correctly or just don't collect it, and then you can fix it along the way. But, you know, I struggle with the notion of, well, you can just leave it broken, you're not looking at it. Um, John, what is your thought on that?

Yeah, well, I mean, you bring up a great point, and that, you know, that doesn't mean that one year from now the new CMO is going to want to see all this stuff, and if it's broken, you're gonna be building it from scratch. So that's yeah, you're you're kind of trapped. It's like, at what point, you know, do you continue to set up measurement? And I think it ultimately you get to a point where you do want to have some fixed rules as far as, hey, I spend X amount of time on campaigns and making things work, and I spend another, you know, block of time on measurement and making things that are working so that you know you don't fall into that trap of like you're just running campaigns all the time and you have no measurement on anything and you're just kind of praying that things work. Um but yeah, it's kind of a a huge ongoing challenge.

The other thing is you can at least, you know, keep thin slicing stuff. Like if, okay, you know, the data from last year is just hosers, never gonna be there. Well, let's start comparing this week to last week, or you know, you know, keep thin slicing down. So at least you've got something to compare against and and see how it goes. And the cool thing with that is you know, you can stretch that out up to a quarter, and then once you start to get to quarterly and seasonal data, it starts to stabilize.

But at least you've got some kind of picture of what's going on. And that's really that's the essence of it. If the data is broken in predictable ways uh from your past, you can export it and and clean it up and and and make it apples to apples. If the data is broken in unpredictable ways or it's just flat out missing, you're you're you're kind of hosed at that point. There's there's not any way to bring it back from the dead.

So at the end of the day, the so what of all of this is that data governance, regardless of what system you're using, is really important to set up from the get-go. And so um, as I was saying in our workshop from a couple weeks ago, Chris, I see Google Analytics 4 as an opportunity to recast those expectations. If you hadn't been using good data governance before, this is a great time to start putting those processes in place because it's a new system, it's an opportunity to have that conversation with your teams, with your agencies, with whoever's working on you know your marketing to say this is the way it needs to be done moving forward. If this is a new system, it's a fresh start, let's get it right this time. Exactly right.

Analytics amnesty. The the so what to me is this. If you want this data and attribution to be correct, you've got to have good governance, right? If you want to be able to look at this tool and say, is what we're doing working, you've got to have the governance in place to make sure the data going into this that feeds this is also correct. Because it's it's computing one-on-one, garbage in, garbage out.

You can't make decisions from broken data. And as awesome as these analytics tools are, they're only as good as the information we put in. So if you wanna hand this to the CMO and say, hey, I want more budget for X, gotta have the data there. And it's the exact reason why we say don't skip over the requirements gathering phase of any project, whether it's introducing a new piece of software or you know putting together a report. That's why the requirements gathering is so critical, because that's where you start to establish those governance rules.

And it doesn't have to be long and laborious. We have a lot of resources on our website that talk about how to accomplish these things. Um, but this is why. This is why skipping to just plugging in the system and saying, okay, new software done isn't gonna cut it. Exactly.

So uh any parting thoughts? Looking at you, John. Dude, you gotta do this. This is everybody fails at this, and this is what you have to do. If you're gonna prove what's working, that this is how you get it done.

Actually, you bring out a really good point, Sean. In one year and 15 days, um, universal analytics turns off, right? The the data stops being collected, and Google Analytics for is the only way to do this. If I guess again, I was on the phone with somebody this morning, they said, Yeah, some of our affiliates are are gonna wait till next year. I said, then they can't get year over year data, right?

If they when when on July 1, 2023, when they finally get off their butts and make the switch, they are starting with one day of analytics data because their old data doesn't come across. So the sooner you do this stuff, the sooner you get your governance in place, the sooner you have good year-over-year data. And that's the short version. Um, if you would like more information about uh the course that we mentioned, uh you can find it at trustinsights.ai slash GA4 course. Uh it is a fairly extensive course.

We put it together, it's like five and a half hours to teach you all this stuff that we just showed, plus so so much more. So please go check that out. Uh, I guess on that note, we will talk to you all next week. Thanks for watching today. Be sure to subscribe to our show wherever you're watching it.

For more resources and to learn more, check out the Trust Insights Podcast at TrustInsights.ai slash TI podcast and a weekly email newsletter at TrustInsights.ai slash newsletter. Got questions about what you saw in today's episode? Join our free analytics for marketers Slack group at TrustInsights.ai slash analytics for marketers. See you next time.