#76 The end of rules-based attribution is nigh!

This week Dan and Dara are back solo this week to talk about the recent news from Google that they are removing most of the rules-based attribution models from Google Analytics 4 and Google Ads. Dara keeps asking if anyone cares or is affected by the change, and Dan gets it, but it not super happy to be without linear attribution!

Enrolment is now open for June’s cohort of the GA4 Immersion 6-week cohort training with early bird pricing for 25% off!

We have an open call to our listeners – if anybody out there is working with Google Ads and is seeing repercussions of this news/update, then let us know and come on the podcast to talk about the marketing aspect!

The announcement from GA4 can be found here, and the one for Google Ads here.

In other news, Dan plays more games and Dara gets some sun!

Follow Measurelab on LinkedIn and on Twitter.

Intro music composed by Confidential – check out their lo-fi beats on Spotify.

If you’re like what we’re doing here, please show some support and leave a rating on Apple, Spotify, or wherever really.

---

Let us know what you think and fill out the Feedback Form, or email podcast@measurelab.co.uk to drop Dan and Dara a message directly.

Quote of the episode from Dan:

“…they’re forcing us to use a model that upweights the Google Stack marketing or the Google marketing platform products, which I understand, I’m not bitter about it, I get it. If I was Google, I’d be doing the same thing.”

Quote of the episode from Dan:

“…if you don’t trust Google, then you won’t be using Google Analytics anyway. So I think everyone, whether you like it or not, has an inherent trust placed in Google to track all your data and do all this modelling and sync it to Google Ads…”

Quote of the episode from Dara:

“In a way it’s amazing how long we got away with being able to look at a cleaner picture of attribution within GA, because one of the benefits, you’ve probably done this too, you would push somebody to use GA data rather than Google Ads data because of the fact that GA included all the other channels and treated them equally with the exception of direct, obviously.”

---

Transcript

The full transcript is below, or you can view the Google Doc.

• Intro [00:15]
• Topic [01:31]
• Wind down [33:00]
• Outro [35:20]

Intro

[00:00:15] Dara: On today’s episode, Dan and I talk about the very recent news that both Google Analytics and Google Ads will be getting rid of some of the rules-based attribution models. Leaving only data-driven and last non-direct click.

[00:00:28] Daniel: So rest in peace, linear, time decay, position based and everything else. But if you want to learn a bit more around Google Analytics, what it has to offer we are currently underway with our first Google Analytics 4 training cohort. This kicked off a couple of weeks ago, it’s a six week program and we’ve got our next cohort running in June. So if you want to join the last opportunity to jump on board and learn Google Analytics 4 with a bunch of peers and professionals and experts, then that’s your opportunity before the big day that Google Analytics gets switched off. So check out measurelab.co.uk/training for more information. And we’re currently running an ‘early bird’ promotion that gives you 25% off while stocks last.

[00:01:05] Dara: Enjoy the show.

[00:01:06] Dara: Hello and welcome back to The Measure Pod, a podcast for analytics and data enthusiasts. I’m Dara, I’m CEO at Measurelab.

[00:01:13] Daniel: And I’m Dan, I’m an analytics consultant and trainer also at Measurelab.[00:01:17] Dara: So it’s just you and I again today, Dan. It feels like it’s been a little while, so we have to come up with something interesting to talk about. So I’m going to turn to you, put the pressure on you to come up with a interesting and engaging and challenging topic for us to cover today.

Topic

[00:01:31] Daniel: Well, luckily Google seems to change something every five minutes, so I don’t have to come up with anything when Google have announced a bunch of changes that I think are going to be the topic of conversation for today. The main one and the biggest one and possibly the biggest change they’ve made since, dare I say, biggest change they’ve made since they’ve decided to deprecate Universal Analytics is that they’re removing a number of the attribution models that are available in GA4.

[00:01:54] Daniel: And now these are attribution models that have been around since the beginning, in a sense. They’ve been around since Universal Analytics. Even in Universal Analytics, we can even create custom attribution models. But finally, they’ve come to the end of their life cycle in Google’s eyes, and they’re going to be switching off a few of the rules based attribution models. Specifically, they’re going to be removing first-click, linear, time decay, and position based attribution, leaving only last-click, last ads preferred click and data-driven attribution to be used by us all.

[00:02:22] Dara: Good old trusty last-click. I could just imagine all of the people who would be truly devastated if they announced the news that they were getting rid of last-click. But at the risk of jumping ahead slightly, we’ll park this but one of the questions, one of the first questions that comes to my mind is, who cares?

[00:02:39] Daniel: This is going to be another classic Dan versus Dara conversation I think. I think I care, I care not just because I cut my teeth in this industry in an attribution platform, or I’ve been using attribution or the variety of attribution models, should I say since I’ve ever been using Google Analytics and even beforehand. The way I’ve always explained attribution when I’m talking to clients or colleagues or people I’m training around Google Analytics, especially when it comes to marketing analytics, is around having access to data. And I think an attribution model is just a lens you can apply on your data to get a different perspective of what’s going on. For me, what I’m feeling like, at least right now initially, is that they’re removing these lenses, which means I’m removing the ability to build a better narrative around what’s going on with my customers around my marketing.

[00:03:23] Dara: To be clear, and let’s maybe, this is probably a good point to maybe zoom out a little bit and go through the detail of what’s been announced. I have to admit, I’m a little hazy on whether I know you think this is being removed even from the model comparison tool, and I was a bit less sure about that because I know it’s being removed from the, the kind of active way you can use it. So from the default selection in GA4, so you’re not going to be able to change the default. It’s going to be data-driven and if you don’t meet the thresholds for data-driven it will default back to last non-direct click. But it was always previously quite useful and this is leading me back to my kind of who cares question.

[00:04:00] Dara: And I know obviously some people really do care, but I think by and large, the majority of businesses using GA are probably still relying on last-click, and at most if they’ve even done this, they might have gone into the model comparison tool just to have a look and compare the different models. Which wasn’t always that useful anyway because of the fact that you’d always get that weird thing where direct would look great in every other model. So it often shows you some slightly weird numbers anyway, but beyond that I’d love to know, and I mean, it’d be amazing to know this, how many GA accounts actually have a default model that isn’t data-driven or last-click. If anyone listening knows that, please let us know.

[00:04:40] Daniel: Yeah, I’m inclined to agree and maybe we’ll put a LinkedIn poll up or something like that. But I think the thing for me is that you’re quite right that there’s two ways of thinking about this. First of all, most people don’t change the defaults, whatever they are, and Google Analytics has defaulted the default attribution model to data-driven attribution. Which in itself, if you don’t hit a certain threshold of data volumes defaults to last-click attribution. So in a sense, we’re using last-click and data-driven out of the box already. And this doesn’t, from what the announcement reads, this is not going to affect the acquisition reports in the reports workspace in GA4. So your traffic acquisition and user acquisition reports, which are in a sense first-click attribution and last-click attribution, they’re going to stay.

[00:05:17] Daniel: What they are removing is the ability to use them as a default model for GA. So you can only use data-driven, I assume they’re just going to make it data-driven and not even give you the choice of last-click. And they’re removing the opportunity to use it in the advertising workspace or the attribution reports they call it. So that is the model comparison reports and the other ones from what it says here. So it does read from the announcement, which we’ll share a link in the show notes. It does read that they’re removing the ability to use anything but last-click or data-driven in all reporting spaces in Google Analytics 4. The only way I can conceive of being able to get like a linear or a first-click back or a time decay model is to be building it ourselves in a sense. Maybe there’s some open source libraries or packages we can use, but using the BigQuery export and doing it ourselves in SQL.

[00:06:00] Dara: I guess I always think whenever I hear that, like that is obviously useful, but not being able to change it within GA is going to mean that you’re going to have that issue, which exists, I guess already anyway, where you maybe have certain people within a business who are using modelled data in BigQuery and then maybe are feeding that into a dashboard. But you’re probably still going to have some people who are going into the GA interface, and then you’ve got that age old question comes up of why did these numbers not match up? So just maybe this is, well, it is going to limit the amount you can do if you do create your own models outside of GA, unless you just bypass the interface completely, which I guess some people will.

[00:06:36] Daniel: Well yeah, but then I mean this poses the question of then, well, why use GA in the first place? I think this is another sort of like nail in the coffin in a sense of Google Analytics being a data product and it’s more of an advertising product now. So Universal Analytics, I’ve always explained Universal Analytics was a data tool with a bolt on advertising marketing module, right? GA4 is a marketing product with a bolt on data module and I think the way that this is going is kind of reinforcing that idea for me at least. You know, that this is going to just focus on data-driven attribution. The only reason I can perceive of that is to better credit digital marketing channels, such as, for example, Google Ads, Display & Video 360 and Search Ads 360, right? So the Google marketing suite.

[00:07:16] Daniel: There’s no other reason, other than computation, it’s not costing Google anything by having these models available, right? It’s just a, you know, as a dropdown in a model comparison report, it’s not doing any harm. So the removal of them is to, in a sense, forcibly encourage people not to use them to look at the data from a perspective of Google and the data-driven attribution model; they don’t share the source code, so we don’t know exactly how it works. But we do know that it upweights Google advertising slightly more than other advertising, right? So in a sense, they’re forcing us to use a model that upweights the Google Stack marketing or the Google marketing platform products, which I understand, I’m not bitter about it, I get it. If I was Google, I’d be doing the same thing. With a lack of visibility across data with things like conversion modelling, data-driven attribution, machine learning modelling, behavioural modelling, you know, where consent is not provided through things like consent mode, it’s all going into this big black box.

[00:08:02] Daniel: And then machine learning happens, and a variety of different levels then out spits a result. And I think that’s the nature and the reality of things nowadays. And I think this is just another one of those inevitable changes where, you know, acquisition reporting and attribution modelling is going to be all machine learning modelling. Again, you know, it’s another layer of machine learning modelling that’s obscuring the underlying data. And yes, we get access to the underlying data. Yes, we can use BigQuery. It’s almost like a lazy answer to everything nowadays. Just use BigQuery, do it yourself, most people won’t, most people don’t know how to, but yeah, that’s still not going to help because you know, Google does so much to embellish the data that we are never going to be able to replicate this stuff in BigQuery no matter how hard we try. There’s lots of people trying really hard just to replicate standard reports in GA, and yet they’re still struggling because the output, the data that Google outputs, the raw data to BigQuery is fundamentally different and misses a lot of the sort of nuance that Google does on top of the data, which they don’t share.

[00:08:55] Dara: Yeah, and again, I would go out on a limb and suggest that the vast majority of users aren’t going to be too effective by this, by this change. You do have your very advanced users who like to use all of the features and will have gone in and changed their default attribution model, created their own rule-based models, etc. But I think most people, that’s not going to be the case. If I’m to be uncharacteristically less cynical, non cynical, you know, Google do have that data. So maybe, maybe they have looked and seen that, you know, not too many people are using, are changing the defaults. So they figure, well, let’s get rid of that option. And the thing with the data-driven is, I always think, it’s the unknown that gets people, isn’t it? Because in theory, at least, using the data-driven attribution modelling is going to be better than any arbitrary kind of rules based or semi data based kind of rules that you come up with based on your own understanding.

[00:09:48] Dara: It’s always going to be limited to some extent. You’re not going to be able to make it as customised, as kind of optimised and fine tuned as you could get with machine learning based attribution modelling. The problem is you don’t know how it’s working and you don’t know if there is some intentional bias in that, where it is upweighting Google’s channels. So it’s probably the unknown that’s going to get a lot of people, but something you said there as well, you know, it’s like when something changes, people go up in arms and it’s usually the, probably the minority people like us who will complain, but before long, to most people this will be the only way it’s ever been. They’ll take for granted that the numbers that Google Analytics says are their conversions will be correct, or to some, you know, correct in inverted commas, and they’ll then maybe have a vague understanding that there’s some machine learning going on behind the scenes.

[00:10:33] Dara: But you’re right, this isn’t something that really, if you want to continue to use the Google stack, then you don’t really have a choice, you’ve just got to accept this, you either stick with good old trusty last-click attribution, or you put your faith in Google’s hands and think, okay, we’ll use the data-driven and hope that it’s being done in a reasonably good way.

[00:10:51] Daniel: Well, for sure, I completely agree and the reason I think I feel strongly about this is because I’ve used them and I’m one of maybe the minority that have used these specifically mourning the parsing of linear attribution, which was my favourite because it’s the my go-to attribution model. It’s easy to, or the easiest one to explain outside of last-click, and it’s something that I used often to kind of demonstrate the inherent biases or the limits of last-click attribution or first-click attribution. So for me, I’m going to miss linear, I’m going to miss those other things. It doesn’t mean I’m not still going to be able to do some of the stuff I can already do. And yes, of course we can go down the path of building ourselves.

[00:11:23] Daniel: I think for me it’s more like, and I think this is where people might feel a bit of an over inflated sense of outrage around this is like, even though I can’t go to your party, I still want to be invited, right? I want to have the option of this thing there, knowing it’s there, and then me deciding not to use it is different for it to be removed entirely. So I think that’s the justification they’ve given for removing this, if I just read it out verbatim because it just doesn’t sound satisfying enough basically. It says:

“These models don’t provide the flexibility needed to adapt to evolving customer journeys. Data-driven attribution uses advanced AI to understand the impact each touchpoint has on our conversion. That’s why we made data-driven attribution, the default attribution model in GA4 and Google Ads. For these reasons, first-click, linear, time decay, and position based attribution across Google Analytics 4 will be going away.”

[00:12:08] Daniel: So in a sense, it’s just saying times are changing, data-driven’s better, we’re going to remove the old stuff. It doesn’t feel like it’s explained it. And I don’t know if we’ve done a good enough job at the beginning of this conversation, at least Dara, just to say that this is a Google Ads thing as well as Google Analytics 4. This is not just a reporting thing, this is going to affect how people manage and optimise their kind of Google Ads campaigns too. It is a broader thing than Google are stopping to support these models, not just in Google Analytics 4, but across the kind of advertising ecosystem as a whole.

[00:12:38] Dara: I kind of like it. Okay it’s a classically, kind of short, to the point, Google kind of help article explanation. You know, it doesn’t go into, it’s almost like the less detail you give, the less opportunity you have to catch your, you know, get yourself caught out or over explain and give away some information you didn’t want to. But this is kind of what I was saying, isn’t it? It’s like if in theory, using machine learning is going to come up with a better end result than you just going in yourself and thinking, I’ll just create some relatively generic kind of rule-based attribution models. The problem again is that you can’t tinker with that in any way. You get very limited visibility on that, but then that’s probably the case with, that’s going to be the case more and more. And you kind of hinted at this earlier, more and more of the day, and we’ve talked about this on this podcast before as well, more and more of the data is becoming modelled, that’s not going to stop.

[00:13:28] Dara: Well, we haven’t reached the end of that journey. The observed data has got gaps left, right, and centre. So those gaps have to be plugged in some way. So maybe this is just another example of that and the fact that you can’t see into that black box. I mean, what would you do if you could see into it?

[00:13:43] Daniel: Probably not a lot.

[00:13:44] Dara: Not a lot. Just look at it and say, yeah, that looks okay.

[00:13:48] Daniel: Yeah, we’ll talk about it on this podcast, and that’ll be an episode done right? They’ve opened the black box, they’ve shut it again. No, but I think just on back on something you just mentioned around this idea that these kind of algorithmic models are always going to be better than rules-based models. I’ve always found that a bit divisive because yes, a machine is going to be better at assigning rules with an unbiased, assuming there’s no bias built in, but an unbiased nature. Like if you ask a paid search marketer what model to use versus a social marketer versus a display marketer, they’re all going to pick a different one that makes their numbers look higher right? And so you kind of remove that aspect of it. The thing that I like about rules based models, and I think this is the thing that often gets overlooked, is that they are fixed. They are rules based, they are static in a sense that the rule that how we attribute email, for example, isn’t going to change next year.

[00:14:31] Daniel: So if I’m doing month or month or year on year reporting, the model of which I’m assessing value has not changed. And so I’m comparing like-for-like, apples-to-apples. The thing about an algorithmic attribution model is the value in the way that I’m rewarding email today is different to next week, which is different to next month and different to next year. So although we’re still looking at a year on year or month on month report, the whole methodology of what we’re looking at is kind of you know, apples-to-oranges, you know, in a sense it’s an evolving thing and it’s never static. I don’t know the frequency of which they update their rules. Maybe weekly or at least it used to be in GA 360 and Universal Analytics. So I think this is the thing is there’s no consistency anymore, like in a sense, everything’s always in flux, in change, evolving. And so data-driven attribution is another one of that where you might see that 50% of your conversions go to email this month. Next month it might be 10% of conversions go to email, but you’ve done exactly the same marketing activity. But the modelling behind the scenes has changed and used a different approach to email marketing, and it does beg another question of like, how quickly can it react to change?

[00:15:30] Daniel: If I introduce a new marketing channel in today, how quickly before that starts to get attributed value, or how long until the machine learns to recognise it and understand the uplift or down lift that this channel makes. Same as if I remove a marketing channel from my marketing mix, like how long before it keeps trying to, you know, like reserve credit for a channel that doesn’t exist anymore. And I think this is the stuff you never have to consider in a rules-based model. It’s just a fixed way of approaching things that will never have these things thrown into consideration.

[00:15:58] Dara: Yeah, I was just thinking as I was listening to you, if it’s totally opening the black box, it’s not going to be useful because you’re not going to know what you’re looking at. But if it could combine it with some kind of like weekly or monthly insights report where it would tell you the reason why email is now getting half the attributed credit that it was before is because of X, Y, or Z. And then have a kind of percentage contribution of those different factors. So it might say, you know, you introduced a new channel and we think the likelihood that affected email’s role in the conversion journey was 70% and then 20% of it was because you changed the landing pages for your emails or whatever. That would be useful, wouldn’t it?

[00:16:37] Dara: Because with what you’re saying, I get it. It’s like you won’t necessarily know what it was that you, you could have changed so many different things because you would think the model wouldn’t change how it’s crediting a channel unless something does genuinely change. So even if your email activity is exactly the same, something else must have changed. Or the model, when it gets updated, it won’t change how it’s crediting email unless something else in that chain has changed. But you don’t know what that is, so if it could kind of also give you a little report each time that the model updates and say, this is why you’re seeing this, or this is the likelihood of why you’re seeing this, then that would be useful, and then maybe people would be okay about it. But it’s not going to tell you anything, it’s just going to give you, it’s just going to output the numbers, and you’re just going to have to trust that they’re reliable.

[00:17:22] Daniel: Well yeah, trust is a huge part of this anyway, and I think if you’re using, if you don’t trust Google, then you won’t be using Google Analytics anyway. So I think everyone, whether you like it or not, has an inherent trust placed in Google to track all your data and do all this modelling and sync it to Google Ads, or you’re advertising through Google Ads. I bet someone’s using some automated system there whether it’s automatic bid optimisations, budget adjustments, performance max campaigns, you know, even looking at some of the audiences, the audience expansions or the kind of the new version of lookalike audiences. Like you’ve got to have a level of trust that Google knows what they’re doing and they’re targeting the right people right? So I think it’s just another one of those.

[00:17:57] Daniel: But just on the subject of Google Ads though. So we talked a lot about Google Analytics 4 and how, I suppose I’ve used it or we’ve used it in the past. Specifically thinking about the Google Ads side, and I think this is something that was announced about a couple of weeks ago and they say in the coming weeks, so I don’t know exactly when this is. But Google Ads, anyone with a Google Ads account might’ve got this email recently, but they’re now moving over to using the Google Analytics 4 attribution model. So, at the moment, if you select something like data-driven attribution or last-click attribution in Google Ads, it’s only going to do that attribution model on top of the data that Google Ads has, which is only Google Ads click and impression data, right? What this update says is that they’re moving over to using the data-driven attribution conversion values from Google Analytics 4.

[00:18:41] Daniel: So this is in a sense, it’s not, I don’t think just a coincidental timing of two features releasing. I think it’s all moving towards one thing, which is we are going to make data-driven attribution the default. And when you use your conversions from Google Analytics 4 in Google Ads, we are going to output or export for Google Ads to use the data-driven attribution credit for Google Ads. So it’s no longer just going to rely on, you know, the last-click data from Google Analytics, which a lot of people didn’t use anyway in Google Ads because it was never as good as the pixel. I think if we look at all of these things kind of in total, holistically, Google Ads is moving over to using Google Analytics 4 conversions. I think all signs are pointing towards that, I would even put money that on in the next year to two years that we are going to see a switch off of the Google Ads pixels and the Floodlight tags as well, just because like everything’s moving over to GA4 and the modelling that it’s doing there, the data-driven distribution modelling that it’s doing there.

[00:19:30] Daniel: So if everything’s moving that way, if we take that on faith that, you know, that might happen. If they’re moving over to using data-driven attribution as the default model in Google Analytics 4 and data-driven attribution credit is going to be exported to Google Ads for optimisation, there’s a lot of things happening at the same time. But from a marketing perspective, what it means is that you’re going to be using different data to credit your campaigns and which feeds the model over there, the optimisation and the, you know, the advertising and the, the bid adjustment models too. So the kind of ROI, the CPA, you know, all of that kind of lovely data’s going to be changing because you’re going to be using different underlying data alongside all these other changes. It’s all data-driven, of course, but you’re now using the data from Google Ads.

[00:20:09] Daniel: All of this is happening in May, by the way. So at the time we’re recording, it’s the beginning of April. And so by May this is going to be for all new accounts set up, it’s going to be this way. And then eventually they’re going to switch off these old attribution models in both platforms by September. So in terms of the timeline, it’s pretty quick, I don’t think that they are. Well, I think that they are very heavily related features, you know?

[00:20:28] Dara: I think you’re right, and that’s a theory you’ve, I can’t remember if you’ve mentioned it on the podcast before, but it’s certainly something you’ve said to me in real life. And I wouldn’t argue with you, it seems like they’re simplifying and they are kind of converging the different products and the focus will be on GA4 which does make, it does make complete sense and goes back again to the point about if you are tied in, shouldn’t say tied in. If you’re using the Google stack, then you’re going to accept this, you’re going to understand that, you know, Google is going to make these decisions that are you know, at times potentially favour them, but why wouldn’t they? It’s their tech at the end of the day. You mentioned people need to have a bit of trust, I think they don’t even need to have that much trust. They need to have some trust obviously, but you don’t have a choice if you’re using Google, and that’s what you’ve always used then the cost to move away, not just financial cost, but of retraining people and learning new systems is so high that people will just accept it. It’s a new change, they’ll accept it and they’ll move on.

[00:21:26] Daniel: This kind of harks back to what we were saying about how you can just use the lazy excuse of just saying, yeah, just do it yourself in BigQuery, right? Another option there is just use a different product. Like when they announced the switch off of Universal Analytics and every analytics vendor came out of the woodwork saying, you know, Google Analytics isn’t dead over here, we’ve replicated the dashboards, we’ve got the thing over here. I think this is going to be another, a bit of catnip for those other analytics vendors that are going to be like, we’ve still got all the attribution models, we’re still doing attribution. Which I, again, I don’t think any of these things are real deterrent for the customers that use Google, because it’s all going to be in the service of the Google ecosystem, which is really huge, right?

[00:22:00] Daniel: I mean, that’s the reality. And obviously it’s free, but nothing’s truly free, like what you were saying is there’s going to be compromise or cost in some way. Either you pay for the product and it is, you know, in a sense non-Google and it’s yours or you get a free product like Google Analytics, but you pay for it in a sense of, you know, a different way in terms of data or using their tailored models that are maybe Google weighted and things like this. I mean, you’ve always got a choice and I would always say assess all options all the time. You know, especially if you’re making a change. You know, assess all options, including not Google Analytics but I think the other reality, and I think this is just another fact of the digital analytics marketing life is that everyone will end up using Google Analytics in some way anyway, because they’ll run ads through Google or the Google ecosystem.

[00:22:40] Daniel: Just another bit on the whole convergence of all these products is that, you know, remember recently they moved everything over to the Google tag, the one Google tag. And so now you put one tag on your website that does all floodlights, Google Analytics, Google Ads tags, and so you don’t need to have three separate pieces of code anymore, it’s one tag that does it all. So in a sense, now when you implement Google Analytics or Google Ads, you are implementing the same product, you’re implementing the Google Tag. And actually all of that technology is built on Tag Manager anyways. The point is, is that there is no difference in the implementation now between Google Ads, Google Analytics, and Floodlights. And I think what they’re going to do now that everyone’s implementing the same literal code on their website or app. They’re just going to move the behind the scenes stuff and just move it over to one place. So I think, you know, with all this stuff, I think, you know, we know what we’re getting into, we know who we’re buying from here in a sense, right? We know it’s Google, we know Google have got a vested interest in making money, and that’s through advertising so, you know, you can’t blame them.

[00:23:31] Dara: Again, I’m feeling very charitable today, far less cynical than usual. But you know, there’s some benefits I guess, if they tighten up the integrations between the two products, between Google Ads and GA4, then it, you know, at least in theory, it should reduce some of the discrepancies that you see when you’re comparing numbers between the two, it’s less implementation as well. So there are some benefits you know, it’s cleaner, simpler implementation. This should be tighter matching of numbers between the two. Obviously you’ll still have differences between, you know, expected differences between metrics like clicks versus sessions. There are some benefits to this as well and yeah, nothing comes free does it? So you have to accept as a result of that, that it could become a little bit more opaque or you’ll have to accept that some of the numbers may have some bias in them potentially, but when has that ever not been the case.

[00:24:18] Daniel: To be more cynical if you are not going to be Dara. I would say that this is one step closer to the idea that you spend your money through Google to advertise. It’s got an audience, which you can’t see, it then tells you how well it’s doing through models that we can’t access and reports on its own performance in a way that we can’t tell if it’s correct or not. So I think this is a, a classic sense if I use the term, of just them marking their own homework. So I think this is, you know, to be the cynic here, it’s one step closer towards being unable to validate. Not only, you know, things that are already removed, like keywords from SEO or you know, actual audience, people in audiences and things like third party cookies going away and audience matching and lookalikes all disappearing.

[00:24:59] Daniel: And then you’ve got machine learning trusting in Google’s idea of machine learning and use of it that they are spending your money, that people are really clicking on it, that they’re reaching the people that they say they’re reaching, and then we are saying, and then tell me how well they went, how much money did we make from that campaign? All from a perspective of the company that’s serving the ad is doing the full thing, which, you know, as a independent analytics agency that we are at Dara, you know, that’s one of our kind of bread and butter statements is when we go work with clients, it’s like, well, your agency might do that. Your agency might be running your campaigns and then marking their own homework by using sort of attribution models or data points or ways of viewing data that make it look more profitable or valuable than maybe we could provide as an unbiased, agnostic perspective. But in a sense Google’s closing the full loop, they’re closing that loop so that you know it’s going to be very hard for other people to be able to be sort of an objective or kind of third party in that whole system right?

[00:25:53] Dara: Yeah, and you’re bringing my cynicism back out again, which I’m quite happy about. In a way it’s amazing how long we got away with being able to look at a cleaner picture of attribution within GA, because one of the benefits, you’ve probably done this too, you would push somebody to use GA data rather than Google Ads data because of the fact that GA included all the other channels and treated them equally with the exception of direct, obviously. So you could say, well, look, in GA you get to have a clear view and whether you like last-click attribution or not, all channels are being treated the same with the exception of direct. So it was kind of surprising in a way that you could do that in a platform that was basically provided mostly for free as a result of people spending on Google Ads. And now, finally, now that they have the opportunity, I guess, to do it with maybe, maybe they’ve calculated this is the point in time where it’s going to cost the least amount of controversy to do this. And to actually say, look, we’re getting rid of these. You’re going to have data-driven, last-click, that’s it.

[00:26:50] Daniel: It’s going to be interesting to see where the product of Google Analytics goes, sort of post this and what the next change like this will be. What’s the next step towards this kind of unity of all of their product suite or their advertising suite here?

[00:27:03] Dara: One interface.

[00:27:04] Daniel: One interface, yeah god. The thing that we call keep talking about on this is just that they really want to rock the boat of Google Ads. Because Google Ads is like 90% of Alphabet’s revenue. It’s like over a hundred million or billion a year or something like crazy like that. So it’s like, do they risk even rocking the boat gently and some of that spilling over the edge? And I think maybe, maybe they do. But I think it’s you know, when you read stories like, Facebook, Meta lost 10 billion last year because of a change iOS rolled out and it’s like, well, if Google had that, it’s going to be a lot bigger than 10 billion because they’re a bigger advertising company and so all of a sudden it’s like, if that’s the evolution of marketing is kind of restriction and lockdown of like tracking and third party pixels and things like that. If Google’s going all in on machine learning and AI to solve these gaps, then why the hell would they not move everything over to it? Because some of their sort of material collateral is, they’ve quote unquote solved the privacy issue in Google Analytics 4, so why would they not use that solved version of the truth in all their advertising stack, and they get a bit more control there.

[00:28:04] Dara: Is there an exploration technique you can use to look at different touchpoints in a conversion journey, or is that information just not going to be available anymore? Unless you use the raw export to BigQuery.

[00:28:15] Daniel: So there’s no exploration technique as such. I suppose you could look at funnels and things like that, or you could look at the user explorer, which gives you a bit of a, like a CRM record which is a bit crap to be fair. But there is the conversion partners report in the advertising workspace that does give you, that gives you every interaction bar direct of course, leading up to a point of conversion. So I think you’ve still got that data, you’ve still got the visibility into that, but it’s still not going to be, it depends what you want to do with it, because it’s quite a micro level report. Like you’re going to look at every unique journey and quantify them. It’s not going to, I don’t know if it’s going to give you what you need.

[00:28:47] Dara: Well that’s the one actually, that’s what I was thinking of. In the past at least that was useful if you wanted to see how often a channel played a part if you didn’t really care too much about what the part was. How many conversions has this channel contributed in some way too. That was useful, I assumed wrongly, thankfully, that was getting removed as well.

[00:29:08] Daniel: No, so all of those reports will stay, but the attribution model that it’s using is going to be removed except for data-driven. And I think this is the key part. The one thing I would say on that, and I know exactly what you mean because we are looking at the contribution attribution, right? It’s the contributed value of like total journeys it’s kind of existed in. The one thing that GA4 doesn’t allow you to do which Universal did, which I find really annoying, is that you can’t do a filter. You can’t search that to just show you all journeys with a specific channel, and you can’t just search and get the number. It’s so annoying, it’s so frustrating.

[00:29:36] Dara: Back to my first question. How big of a deal is this? How much should people care? And how much do people care? Have you, are you in the thick of it? Are you part of any groups of rebels complaining online about this? Or are you just thinking, you know what, this is fine, it’ll happen not a big deal.

[00:29:52] Daniel: I’m in the camp that it is just going to, it’s inevitable. We don’t have a say, this is not a democracy right. Google has decided and told us that something’s happening. I’m in the camp of like, whatever, sure, another change. I have to update some training content and collateral that I’ve made, you know, that kind of stuff that I’m a little bit peeved about, but it’s just inevitable that things like that will happen. I think it’s been a bit of an equal split, some people welcome the change, some people are against the change. But a lot of the circles that I’m reading through or listening is analytics circles, and I think what’s going to be interesting for me is going to be the marketing circles and wondering the impact that this is going to have alongside that other change of using the GA data in Google Ads. So I think it’s a bit of a split bag at the moment, you know, there’s I would say 50/50 just to be, you know, safe. I’m just going to say it’s a 50/50 split in terms of sentiment. I will miss it, but I also appreciate, you know, or understand it a little bit. You know, just say like, you know, it is what it is, it’s happening regardless. But like I said, I’m going to be interested to see where the marketers, the search marketers, the display market, anyone using Google Ads, I’d love to get their opinion on this.

[00:30:49] Dara: That’s a really fair point actually because I guess mainly why I’m kind of asking that question in a bit of a flippant way, because it’s maybe something that I don’t always need to worry too much about. And often the data coming from GA has been the standard last non click attribution model. But the change affecting Ads is bigger, because what’s going to happen to your performance if you, so you either make the change in advance, you accept that it’s going to change. So what did you say, you said from May I think new properties, and I guess that would apply to new ads accounts, wouldn’t let you change the default. Then from September, all accounts will lose the option to select any of those other models. So if you’ve been bidding based on an attribution model that’s going to get taken away, that could mess up your whole account temporarily while it all readjusts itself. That would be a reason to care?

[00:31:40] Daniel: Well yeah, if you are currently got a marketing or bidding strategy based on something that’s not data-driven, then I would, I would be kind of, you know, there’ll be a sweat drop running down my head and just thinking about how much work have I got to do to rebalance this portfolio of ads, and especially on some of these bigger ads accounts. So I would be moving down that way, but like I say, combining this with the other change I think maybe you could do two birds, one stone, and not have to redo the whole thing twice. But the move over to using the kind of data-driven export from Google Analytics, not the data-driven attribution from Google Ads, I think that’s going to be the next step. I will tell anyone right now, it doesn’t matter who you are, if you are not exporting your GA4 conversions into Google Ads, you need to start right now. Even if they’re secondary conversions purely to run in parallel alongside your Google Ads pixel, you know, while it’s still there you know, all that kind of stuff. But I would always in a sense, be doubling up. I’ll be doing the Google Analytics export to Google Ads for conversions, and I might even be using that already, but I’d at least have it ready to go so that it’s not a shock or a surprise to you, so you’re not kind of like worried about that change.

[00:32:39] Daniel: So yeah, maybe do two at once but I would be, I mean, for me, it removes a tool in my tool belt when I’m doing marketing analytics reporting, budget adjustment, you know, kind of end of campaign analysis, reporting, things like that. I’m not spending tens or hundreds of thousands of dollars a month on someone’s account that’s now about to potentially have to change. You know, I don’t have that level of dependency on these right now.

Wind down

[00:33:00] Dara: And there’s an open call to our listeners, if anybody out there is working in the kind of Google Ad space and is seeing repercussions of this that maybe we’re not paying enough attention to, then let us know and if you want to come on and chat to us about it, then you can do that as well. Okay, I think that’s all we can really say about this for now, Dan. We’ve given our take on it and given people some things to think about, but I guess we just kind of like with a lot of these changes we continue to kind of watch and see what the fallout is like. So it’s been a while, I think I mentioned this at the start. It’s been a while since it’s been just you and I, so we’ve been off the hook for our wind downs for a while. So that should have given you plenty of time to do something interesting. So what have you been up to, to kind of switch off from work lately?

[00:33:45] Daniel: Well, I tell you what Dara, I bought a Steam Deck not too long ago. So this is a handheld gaming PC. And I think I may have mentioned this before, but I’ve been playing this lovely game called Sable, and it’s made by two people in London, so it’s just a two person outfit. And they made this beautiful sort of narrative adventure puzzle game, and I can’t recommend it enough. It’s a non-combat game, and it’s about solving puzzles and exploring and uncovering this narrative around why you’re there and collecting things and doing other things. It’s very stylized, very artistic, and I give it a double thumbs up, Dan’s double thumbs up. So if anyone’s out there and thinking of something to play next, something that’s chill, relaxing, and maybe 10 to 20 hours long, which is a real big positive, rather than these 100 to 200 hour epics that I like to play to, then check out Sable.

[00:34:32] Dara: My mind, even though you told me about the Steam Deck when you said it, I thought about some kind of steam based contraption, for pressing your trousers or something like that. Like an old school iron type, you know, are the ones they have in dry cleaners.

[00:34:46] Daniel: Like a trouser press.

[00:34:47] Dara: Yeah exactly, a trouser press yeah. So it just goes to kind of, well reiterate how little I know about gaming.

[00:34:53] Daniel: What about you, Dara? What have you been doing to wind down?

[00:34:55] Dara: So, I’ve been on holiday at least at the time of recording. So I’m only just back, I got back two nights ago, we went to Tenerife for a bit of sunshine. Lots of outdoor activity, lots of hiking, time in the sea, it was really nice.

[00:35:09] Daniel: Oh, beautiful.

[00:35:09] Dara: And very little time on the screen, which is always great. A true wind down.

[00:35:16] Daniel: Yeah, a true wind down, an escape, amazing.

Outro

Dara: That’s it for this week, to hear more from me and Dan on GA4 and other analytics related topics, all our previous episodes are available in our archive at measurelab.co.uk/podcast. Or you can simply use whatever app you’re using right now to listen to this, to go back and listen to previous episode.

Daniel: And if you want to suggest a topic for something me and Dara should be talking about, or if you want to suggest a guest who we should be talking to, there’s a Google Form in the show notes that you can fill out and leave us a note. Or alternatively, you can just email us at podcast@measurelab.co.uk to get in touch with us both directly.

Dara: Our theme is from Confidential, you can find a link to their music in the show notes. So on behalf of Dan and I, thanks for listening. See you next time.