There are some things in Marketing you just accept as “given” at some point in your career. I’ve been thinking a lot about some of them lately and I’d love to talk about one of these “given truths” today. But let’s start at a very recent obsession of mine:
Pilates!
I actually got Tayfluenced and downloaded the Blogilates App a while ago, since, you know, healthy lifestyle and wanting to get fit and so on. Cassey, the founder of the Blogilates App, is amazing. Even on the very small smartphone screen I’m staring at to do the pilates videos the app provides, she has such a presence and always manages to push me through the hardest minutes of these workouts.
Needless to say I had to buy leggings from her fashion brand, Popflex, too, right?
Well, that was before Taylor Swift wore three of her fits in her latest music video. Which made me realize how cute the skirts were. So I bought some of them, too. And I love them. They are comfortable, they got pockets, they are twirly, I feel pretty in them, I wouldn’t wear anything else, ever, if I could.
At that point, I’m fully hooked on this brand. So obviously I follow them on Instagram. And when they posted a cute video of their backpack, I thought to myself “hmmmm, I could use a new backpack, I’ll be in Hamburg for a couple days in 2 weeks and I actually don’t have a weekend-sized travel-backpack anymore”. So, off to Popflex I went, but I just wasn’t ready to pay 156 Euros, 20 Euros of Shipping and a couple bucks in taxes for a backpack I’ll only use for a handful of trips per year.
But still, I convinced myself to be in desperate need of a backpack for this upcoming trip. So, off to Google I went. A brand search brought me straight to the Amazon Store of Gotbag, as I’ve been eyeing one of these for a while. Then I remembered the tiny shitstorm they once got because their plastic-ocean-saving-advertisement-story might or might not be 100% true. So I ended up buying a very regular, very good-looking, very spacious backpack from a wholly different brand.
With that in mind, let’s talk about Attribution. Who should get the credit for my backpack sale?
What even is Attribution?
The term “attribution” originally doesn’t exclusively belong to marketing, it simply means that you match a certain value to a certain event. In marketing, we’re using the term “attribution” whenever we talk about some kind of conversion, could be a sale, a contact form being sent or simply a newsletter subscription, that needs to be credited to one or multiple marketing campaigns or channels.
So, in one simple sentence, attribution tries to answer the question of “who gets the credit?”.
In coming: Attribution Models.
You could look at my example backpack-sale or maybe even just recap your own shopping behavior and you will notice: It’s not that simple to match a conversion to a specific marketing campaign.
This is where attribution models come in. With attribution models we try to determine the impact of each touchpoint on a sale and therefore how to fairly distribute the credit for any given conversion.
Over the last years, the marketing bubble went kind of crazy with that, there are several tools (we’ll be talking about these later) and dozens of models with different levels of complexity, starting at the very basic regular last-click ending with blackboxed AI models no one really understands.
Five and a Half common and popular Attribution Models
Before talking about the sense and nonsense of attribution and sharing some actual numbers, let’s have a look at the most common models out there.
Last Click / Last Interaction
In this model, 100% of the conversion is attributed to the last click before the purchase. This model is suitable for products that don’t require a long information phase before buying.
Example: Low-priced products like T-shirts or socks.
Last Indirect Click (which is basically last click, just different)
The „last indirect click“ model used to be the most commonly used for a whole while, as it was the default attribution model in Universal Analytics. As the name suggests, the conversion is attributed to the last indirect click – meaning the last click that wasn’t a direct entry of the URL in the browser.
If the user visits the website directly for the conversion and completes the purchase, this model assumes that the preceding marketing campaign was the actual driver of the conversion.
First Click
In the „First Click“ model, the purchase is attributed to the first point of contact. This model is particularly useful for complex, explanation-heavy, or high-priced products or services.
Example: Buying a car, booking a trip.
Linear Attribution
In a linear attribution model, all touchpoints in the customer journey are considered equally important. The conversion is thus distributed proportionally across all contact points.
A linear model can be used when marketing efforts are designed to guide the user consistently throughout the entire decision-making process. However, in practice, this model is rarely used.
Time Decay Attribution
Similar to the linear model, the time decay model distributes the conversion across various touchpoints. However, the distribution is not linear but instead exponentially increases toward the last touchpoint.
In Google Analytics, the half-life of the time decay model is 7 days, with a maximum attribution window of 30 days. Therefore, time decay attribution is especially well-suited for short-term campaigns.
Position-Based Attribution
Here, a conversion is distributed across different contacts. Unlike the previous models, position-based attribution is very flexible: companies can decide for themselves which touchpoints should be given more importance.
I first started this sentence with “these models are great”, but honestly, they are not. They’ve been working for a pretty long while because they are easy to understand, easy to implement (you had to do literally nothing to implement them because they used to be standard in all marketing and analytics platforms) and they gave us numbers to report to our stakeholders, which is what we all needed at some point.
But do these models actually capture which campaign or channel was responsible for the sale? Not necessarily. You just decide that the last touch is the most important one. Is that correct? Nobody knows.
Data driven Attribution Models
So, the existing models were not exactly great. Customer Journeys are getting more complex every day. Still, the desire to attribute your success to a specific channel or campaign is as high as ever. Which is why new solutions were developed – solutions that are more “data driven”.
Here’s an Example: At first glance, it looks like Ad A usually leads to conversions. But when you dig deeper into the analysis, you notice that the conversion rate is higher when users first click on Ad B before clicking on Ad A. In a data-driven attribution model, a portion of the conversion is also attributed to Ad B in this case.
Sounds complicated? Don’t worry, the calculation is automated and oftentimes AI-powered.
Advantages of Data-Driven Attribution:
- It seemingly provides a more realistic picture than standard attribution models.
- It just sounds pretty cool.
Disadvantages of Data-Driven Attribution:
- A relatively large amount of data is required for it to work properly.
- No transparency into the allocation – you have to blindly trust third parties like Google.
We’re four pages in and as of now, the options we’re having is “trusting a blackbox with attribution” and “just deciding which step matters most and going for it, regardless of how true and correct this is”.
Needless to say that both of these options pretty much suck.
Which Attribution Model does Google Analytics use?
If everything sucks, it’s worth looking at how the biggest players in the industry do it. Cue Google Analytics 4 – I mean, they must have figured it out, right?
Haha. Ha. Ha. Well.
I won’t even talk about that in detail, but there’s a lot of mix and match and First and Data Driven and different kinds of source-medium-assignation in different kinds of scopes, and while I think most of it makes sense once you really thought about it, I also think it shouldn’t be as confusing as it is.
Maybe I’ll write my own blog post about all of this one day, but honestly, Charles Farina, one of the smartest people in the industry already talked about all of this in detail in his excellent blog post. It’s a very good read, so if you’re interested in attribution and using GA4 you should really jump over (in a new tab, so you can come back and finish reading this one).
I’m looking for the option to change attribution models in GA4
We all are.
It got discontinued. 🥲
WHAT? How can I compare different attribution models then?
By default, at least if you’re using GA4, you can’t.
There are some third party tools, though, that you could use to compare different attribution models, namely for example Attributy, Tracify, TrippleWhale or, my favorite amongst them, getKlar.
If you don’t want to invest in a third party tool (and I’ll tell you why you might not want to do that, even though there are some okay-ish tools in the market), you can still use the Big Query Export of your GA4 Account and simulate and compare the different models in your own data.
Here’s how you can do this:
Attribution Models in Google Big Query
Joke’s on all of us, Big Query itself doesn’t use any kind of attribution model. In that case, it’s a good thing, though.
If you want to see the comparison between First Interaction and Last Touch, it’s pretty easy, as each Event comes with a User based Source Information (“First user”), an Event Scoped one (“last touch”) and, as Simo Ahava pointed out on Linkedin, finally also a Session Scoped one.
So, there’s a lot you can see and analyze in Big Query, even without going into deeper analysis. If you want more, you can also do multi-touch or position-based attribution modeling with your Big Query Data.
If you want to try that, here’s a very good step-by-step-guide written by Johan van de Werken.
What Attribution Model is the best one?
I wish this was an easy one to answer, because the “Which attribution model should I use?” is a question I’ve been asked a lot.
But honestly, I can’t tell you – at least not in a blog post like that. Which attribution model to use depends on a whole bunch of factors, for example:
- How many and which channels are included in your marketing plan?
- How pricey is your product?
- How complicated is your product?
This vastly influences the choice of the right attribution model as well as the right lookback window.
Limits of Attribution
So, we’ve been talking a lot about how to do attribution, which models to use, how to do them. As of now, there are basically two questions left with the first one being:
“If I choose the best model, UTM-Tag all my channels properly, maybe even buy a third party tool for attribution… Will this finally and correctly tell me the impact of each marketing campaign and teach my on how to allocate my budgets?”
And, as sorry as I am to tell you, the answer is “no”.
There are a multitude of things we can not properly capture in any analytics tool – and therefore not take into account in attribution, for example:
- Views and Inspiration: While there are some campaigns and channels with very wild recommendations and standards around view-attribution (looking at you, Youtube), a standard multi-channel-approach in GA4 or another analytics tool can’t quite work based on platform views, as the analytics tool just doesn’t know about your views.
- In-App-Browsers: This is something very few people actually notice or think about in their daily life, still it’s one of the biggest challenges for tracking: If you’re clicking on an Instagram Ad, in most cases Instagram won’t lead you into the system’s browser, the Landing Page will be opened in an In-App-Browser instead. Once you close that In-App-Browser, the Session is closed and the Cookies are gone.
So, if you, like me, don’t buy right away, but waste all your mental capacity for thinking about that god damn gitti nail polish you’ve seen in that one instagram ad that’s just too expensive to buy, then start to girlmath it for hours and maybe days until you finally talked yourself into believing that you. can. not. survive. without. this. dark plum. nail polish, and you finally come back in your regular browser and buy that very expensive nail polish, you will probably be messing up the attribution for this online shop, because the sale will go to whichever channel come back with and nobody will even see that you obsessed for hours and days based on a social ad you clicked and landed in an in-app-browser.
That being said, I think I found a new favorite gitti nail polish color, because I absentmindedly clicked on the shop when writing this article.
- Consent: This one is easy – you can’t track it, you can’t attribute it.
- Cookies: Same here – if the cookie is gone, you can’t attribute it.
So, however hard we’re trying, attribution will always be a guessing game, at least to a certain extent. Which, on the one hand, really sucks. I get that. On the other hand, though, I’m really glad to be living in a country where consent matters and I’d always put consent before attribution.
Knowing all of what doesn’t work: Is Attribution even worth it?
Let’s quickly come back to the example I mentioned in the middle of this article: How should my nail polish sale get attributed? And my second one? And my third one? And maybe the fourth one I’ll be making as a little treat as soon as this post is published?
I’ll quickly recap my customer journey here:
So.. Who gets the sale?
With first touch, it would be the Insta Ad that actually lead to all the sales. At least in a perfect world. There’s this in-app-browser problematic, though, so it would be the organic channel of my first brand search.
With last touch, it would also be organic and another brand search. Oh well.
Position Based or Time Decay, we’d have a bit organic, a bit paid search and – again, in a perfect world, also influencer and the insta ads. I only viewed the influencer story, never clicked because I already knew the product and the brand. So the Influencer Touchpoint would never even be part of the journey in the attribution report. As well as the other Insta Ads, because of the In-App-Browsers. Obviously, I’m also leaving out the girlmath-part that actually convinced me that it might be okay to spend that amount for a simple nail polish. And Codecheck, which was a very important part of the journey and the only possible thing that could have prevented me from buying at that point, doesn’t even lead back to the Gitti Page, so I didn’t click there, either. No touchpoint for Attribution.
That being said… Who gets all the follow-up-sales that happened months later?
I guess what I’m trying to say is: It’s kind of pointless to try to attribute one or even multiple sales when so many touchpoints are missing.
So, no, attribution might not be worth it, at least not in the way most marketers would love to use and leverage it. In my opinion.
I would still recommend experimenting with different attribution models, comparing them, getting comfortable with the data you’re gathering, learning about the part of the customer journey you can track and control and also learning about what you can’t track.
In short: Take some time to get to know your users better – with all the data points you have on hand, think analytics, but also think beyond analytics. And, please, don’t let the hunt for the perfect attribution model cause you sleepless nights.
Love, M.
P.S.: To answer the initial question: The backpack-sale should be last click.
Not that anyone cared, but here’s the explanation:
While I love popflex and they gave me the initial inspiration for checking out backpacks, I would have never bought one of these – also, I would have gotten the inspiration when trying to pack my bags for the upcoming trip and noticing that I don’t have any backpack. So it’s not popflex credit in that case. It’s also not Gotbag, because again, even though I clicked, buying was never an option. With the last touchpoint (directly into the amazon app and searching for my new backpack there) I found the right product (looks amazing, meets all the criteria) for the right price with the right service (free delivery, free returns).
Therefore, yep, this one was a last click because the other two touchpoints did not influence the decision to buy the backpack I bought.