Which Attribution Model To Use
-Eric Bergman, Paid Media Specialist
Many times I have been asked, “What is attribution?” or “Why should I care about attribution?” My response is always, “Do you want to sell more, or get more patients, etc.?”. Of course, they do, and that is why they should care. As far as answering what attribution is, well, that one is simple but complex at the same time. Attribution is how credit is given to the desired action. That’s the simple answer. Have a good day.
Of course, where the complexity comes into the picture is answering follow up questions like “what does that mean?”, “Why are there different models?” and “which is best, and which attribution model to use?”. Selecting and utilizing the correct attribution modeling in Google Ads and Facebook for your company or project can take you from an okay marketer to an okay marketer who looks like a superstar. Before we get into the weeds of which attribution model is best for you to use, we need to start at the beginning.
Attribution was first explained to me using the following analogy, and I’ve found it to be the easiest way to understand and explain attribution.
There is a seminar I am giving with 100 people in attendance. The chairs are lined up in a 10 x 10 pattern. As the lecture starts, I call out to the person in the very back right corner. I tell them that under their chair is a basket and ask them to pull it out. Once they’ve done that, I let everyone in the audience know that I have placed a token under each chair and ask them all to pull it out.
Now, everyone has the token in hand. I instruct the person in the back right corner to put their token into the basket and pass it around. As you receive the basket, drop your token in, and then give the basket to the next person. As this is going on, I continue my lecture. Once the basket has made it around the audience, it ends up in the front left corner with 100 tokens in it. At this time, I stop my lecture and go over to the individual who has the basket overflowing with tokens. They hand me the basket. I thank them for the basket of these lovely tokens, and I give them a $100 bill for it. That person is very excited and happy, but everyone else is not. The person in the back right corner asks, “What about me? I got the basket and started the whole thing off.” You see, in this example, I don’t care about the basket’s journey (the customer journey) I only care about who had the basket last and who gave it to me.
This moment is where attribution comes into play. What part of that customer journey is most important to you? In this example, everything that came before the handing of the basket (the last click) was ignored. But what happens to those other 99 touchpoints? Do they have value for you? Would you value the person in the back right corner? This person not only got the ball rolling but also contributed a significant part, the basket.
As business owners and/or marketers, we can tend to get so caught up in the results that we ignore the journey that took place to get there. Humans want gratification and give all the value to that sale, that patient, that lead. But, would we have those things without the person in the back right corner? Would we have a basket overflowing with trinkets if it wasn’t for all 100 people? Would we even have that basket?
To me, attribution isn’t just how credit is given to the desired action; it’s following the footsteps of that customer’s journey. How did they get to me? Where did they find me? How many times did they look for me or see me? How do I get more people like them? It helps me visualize all of these things. Now, I am not saying last-click attribution doesn’t have a time and a place, because it does. But that’s why we have to understand what is relevant to you and your company. Also, we need to know what touchpoints at what times have value for you.
To help you understand and decide which attribution model to use in the Google Analytics attribution model and on the Facebook attribution model, let’s take a look at all of the options. To start, we will dive into Google attribution. There are currently six options: Data-Driven, Last Click, First Click, Linear, Time Decay, and Position-Based.
Google Analytics Attribution Model
The default setting in the Google Ads attribution model is ‘Last Click.’ As we saw above, Last Click attribution gives 100% of the credit for the conversion or wanted action to the very last click. So it doesn’t provide any credence to what keyword or click brought that customer to you or even how long their journey was! Last Click is best for items with immediate purchase times where your customer does not do much shopping around, or when used for a specific branded campaign.
The opposite would be First Click, whereas you’ve probably already guessed – 100% credit goes to the first click. In this variation on the analogy, I would have given the $100 to the person in the back left corner. I see First Click used most at the start of a new digital or brand awareness campaigns, where we have a very high level, non-branded top of funnel keywords.
Both models only show us one touchpoint and are not intended to follow the customer path. I would advise you only to use these in very strategic times, if ever. At the same time, I would caution the use of these, especially if you are using conversions when optimizing your account. I once had a client who was getting about 1000 conversions a week. His account was set to Last Click attribution, and he was doing keyword optimization. He noticed he had a handful of keywords that were getting a ton of clicks and spend, but showed no conversions. So he decided that to maximize his conversions and stretch his budget further, he would turn those keywords off because he was not getting any conversion from them. A few weeks after he made these changes, he called me in a panic. Conversions had tanked, dropping from 1000 a week to under 400 a week! He was at a loss and could not figure it out. We started digging into his account. We realized the keywords he turned off – while uncredited for the conversion because they were not last click – were very instrumental in bringing customers in, and keeping them on the journey. Ooops. By turning them off, he basically cut the head off of his account. We turned the keywords back on and changed his attribution model. After a week or so, his account had rebounded. From there, he was able to make much better decisions when it came to optimizations. So once again, I warn you to use Last and First Click models carefully and only at certain times.
When we move onto Linear, we’re starting to see the customer journey. With Linear, we divide 100% of the credit evenly to all touchpoints. So if you have four touchpoints, each gets attributed 25% towards that desired action. It’s a good starting point if we’re unsure what touchpoints we want to focus on or which holds the most value for us. Linear shows us all touchpoints in the journey. Use Linear when you want to see the essential touchpoints with giving no value ranking to any touchpoint.
Time Decay is a great model to use when you have a longer sales cycle or when you are marketing towards a sense of urgency. Time decay favors touchpoints closest to the conversion based on how recent the point was to the actual conversion. Say there are five touch points strung out over four days. The touchpoint closest to the conversion would get the most significant percentage of influence on the conversion, and the first click would get the smallest.
Out of the first five Google analytics attribution models, I prefer Position Based attribution. In this model, the first click and the last click each receive 40% of the conversion value. The remaining 20% gets divided evenly among the other touchpoints. I consider the two most essential touchpoints as ‘what brought them to me’ and ‘what sealed the deal.’ Without knowing how they got to me, how can I properly optimize to get more? Without knowing what closed the deal, how can I make sure I am driving the right people down my funnel? Out of the first five models, I use this one 98% of the time.
THE GOLDEN CHILD – THE CREAM OF THE CROP – is Data-Driven attribution. This model can be available on every account IF you have enough conversion data. There’s no hard line for what that number is because there are many factors. These include but are not limited to the industry, types of conversions tracked, etc. BUT, when your account becomes eligible for the Data-Driven model, USE it. This model uses machine learning to evaluate all converting and non-converting paths and gives out the credit for each touchpoint. This model looks deep into the journey and considers the number of times that person interacted with the ads, the ad creative, order of exposures to the ads, as well as other factors like keywords and clicks.
As you can see, it’s a simple yet complex answer when discussing Google Analytics Attribution models. There is no one size fits all, and each one has its own time and place. The critical part to remember is that you need to know your goals, and what parts of the customer’s journey have value to you. But in the end, if you know what those are, and use the best model for your account, you can optimize with the best of us and really drive success.
Facebook Analytics Attribution Model
Facebook attribution is a different beast altogether. Yes, both Google and Facebook have attribution modeling, and yes, there are similarities. But, there are several significant differences, as well.
Facebook has seven different options (even-credit, first click or visit, first touch, last click or visit, last touch, positional, time decay, and data-driven). Still, before we look at the different models, we need to take a few moments and discuss the window/click windows that will play a role in each one.
The window that you choose determines which touchpoints will be considered by the model. Only touchpoints within the selected windows will be used for the attribution. The windows you can choose are 1, 7, 14, 28, 30, and 90-day windows. This window sets the length that the attribution model is considering, and you will use different windows depending on the sale cycle for your business. The smallest window you can select is a 1-day click/1-day impression. Which means it will only include clicks and views that happened in the day before the conversion event. When you first set up your account, the attribution model defaults to a 28-day click and visit/1-day impression. This is the recommended starting point if you are unsure of what you should be using.
Similar to Google, Facebook’s First Touch Attribution model will only look at a single touchpoint to assign 100% of the credit. If there was no click or visit, then it will default to the impression and give it 100% of that conversion. If the path has an impression that’s followed by a click and finally a visit, the click will get 100% of the conversion, and everything else ignored. When there is an impression, click, and a visit a short time later, the click and visit would be grouped, then counted as the same touchpoint and given 100% of the credit.
First Click or Visit, is very similar to First Touch but only considers clicks and visits and ignores all impressions or multiple clicks and visits. Use it when you have a longer attribution window and want to understand the value of the first click or visit. This model simplifies the whole customer path. If you really want to understand the customer’s path as an entire journey, I would use a different attribution model. This First Click or Visit model will give 100% credit for the conversion to the first click or visit only. If the click and visit happened within 1 minute, they would be grouped and counted as one.
On Facebook, even-credit is just like the Linear model in Google. Every touchpoint from the first impression to the conversion is considered and evenly credited. It doesn’t matter if the touchpoint is an impression, click, or visit. This model shows you the whole customer journey but does not indicate if one touchpoint is more important than another. So if you have three impressions, one click, and one visit, each of the five touchpoints would be given 20% credit towards that conversion. If the click and visit happen within 60 seconds of each other, they are grouped, counted, and credited as a single touchpoint. So in the same example – but with the click and visit happening at 45 seconds of each other – each of the four touch points would get 25% credit instead.
Last click or visit and last touch, are mirror images of first click or visit and first touch. Once again, these settings are only going to show us the last action, and assign it 100% of the credit. All four of these have their time and place – very situational – and only on particular accounts or certain business verticals.
Positional modeling gives the first and last touch a specific percent of the conversion, then splits the remaining credit up among all the other touchpoints. Unlike Google, you can use two different configurations when selecting positional modeling. You can choose from a 40/40/20 split as Google uses, or you can do a 30/30/40 split. That’s where the first and last touchpoints get 30% each, and the remaining 40% gets split up evenly. Both will give you the whole customer path and showcase the first and last clicks (those touchpoints that brought people to you and the touchpoints that sealed the deal on the conversion). These touchpoints can be impressions, clicks and/or views.
Time decay also allows you to see the whole customer journey, but it gives more weight to touchpoints closer to the conversion. In this model, the percent of the conversion given is divided in half after set time amounts. There are two different configurations you can select with Facebook’s time decay model (1-day and 7-day half-life). What that means is if you were using the 1-day half-life, and the most recent touchpoint happened a day ago, that touchpoint would get 50% of the credit, but if it were two days ago, it would only receive 25 percent.
Last, we have the data-driven model. This is a statistical model where Facebook gives incremental credit to every touchpoint based on the algorithm that Facebook has created. It does this by estimating the impact of each touchpoint had on the overall conversion. Like Google, this is Facebook’s golden child of attribution modeling, and you should always consider using it when available.
Wow, that was a long journey for us to take together! As I said at the beginning, attribution modeling is both very simple and very complex at the same time. When trying to understand your clients and how they get from point A to conversion, understanding and selecting the correct attribution model is essential. Each model has its own time and place and will work better for different business verticals and company goals. So before you select one, or just use the default setting, take your time, evaluate your goals. Use the attribution modeling in Google Ads and Facebook that makes the most sense to reach them. And if you’re still not sure, give Serendipit a call.