Every week a different member of our agency gives a presentation to the full team on an interesting digital marketing topic, and our Ad Platform Manager, Dubravka, recently gave a great presentation on “attribution modeling” that we wanted to share, because this is a topic that we discuss with clients on a regular basis. You will find her 17-slide PPT deck below, and here are the eight major things you need to know to introduce yourself to attribution modeling:
The Eight Things You Need To Know About Attribution Modeling
- First, here is the definition of an “attribution model”: an attribution model is the rule/rules that determines how credit is assigned to the touch points along the conversion path
- Second, it is all about the “conversion path.” Many marketers follow a “Rule of 7,” which states that a prospect needs to see a marketing message 7x before they turn into a customer. There are many studies that show different numbers, but the bottom line is that the number is always going to be greater than 1, which is why attribution modeling is necessary.
- Third, you need to understand that “conversion tracking” is a completely separate thing. Conversion tracking is the definition of a “conversion” itself — a purchase, a lead, an enewsletter subscriber. Once you define the conversion, you need to know which type of ad or channel preceded the conversion, and that is where attribution modeling comes in.
- Fourth, it helps to think about several common scenarios that lead to attribution issues — the mobile vs. desktop issue, and the non-brand vs. brand issue. The mobile vs. desktop issue is where you see lots of traffic originating on mobile devices, but all the conversion are occurring on desktop. This type of situation results in the potentially false conclusion that desktop is more valuable than mobile, when in reality the sales are driven by mobile research followed by desktop purchases. If you remove the mobile traffic, you may see a decline in desktop conversions. This is similar to the second scenario, the non-brand vs. brand issue, in which you see non-brand keywords that drive the majority of spend and volume, but you see brand keywords that drive the bulk of the conversions. Again, the issue comes down to non-brand keywords being responsible for product research, and brand keywords being responsible for the final conversion. This scenario is easier to solve with sophisticated attribution modeling, because all searches occur on the same device.
- Fifth, it all started with “last click” attribution. This is very simple, because it literally awards credit to the “last click” that precedes a conversion. This is still the most common attribution model, because many non-professional marketers pick this setting as the default for their campaigns, and it reinforces the problems described in point #4 (the mobile vs. desktop issue, and the brand vs. non-brand issue.)
- Sixth, the most common “simple” attribution model that professional marketers recommend is “first click” attribution, which is exactly the opposite of “last click,” in that it attempts to award ALL credit to the first channel that drove a visitor to your website — the theory being that the first visit is what marketers can control most directly, and therefore it is a big step up from the flawed “last click” model, that favors things like desktop traffic, brand keywords, and direct traffic, all of which are true.
- There is now a proliferation of new attribution models that Dubravka highlights that give a more nuanced understanding of the conversion path that customers follow, including:
- Time-Decay Attribution Model: this is where the touch points that are “closest” in time to the conversion get the most credit. In other words, the more time that passes, the less credit is awarded. This might be good if your customers know you and you don’t need to spend money and time to acquire them, you just need to re- engage them over and over again, then the last interaction is the most important one.
- Linear-Attribution Model: this is where all touchpoints are treated equally.
- Position-based Attribution Model: this is also referred to as a “u-shaped” attribution, because the most credit goes to the first and last touch points, and equal credit is awarded to the middle touch points.
- Data-driven Attribution Model (DDA) — this is the newest, and requires a minimum of 600 conversions and 15,000 clicks over 30-day period, and this attribution model is greyed out in the account if those minimum numbers are not hit. We expect to see much more of this in the future, as it improves.
- Here’s what happens when you change attribution model: Conversions will not be associated with the last click (or whatever your previous attribution model was), and instead, fractional conversion credit will be redistributed to previous days. There is no loss in conversions, just a redistribution. Over time, conversion credit on the day of the change will normalize as it will receive fractional conversion credit from future conversions.
- Which attribution model is best? This depends on: the resources and time you have for data analysis, your business model, and also the “anticipated” journey customers will follow as they make purchase decisions. We always recommend collecting as much “useful” data as possible, with the emphasis on “useful.” If you simply collect as much data as possible, you will build up so much noise that you won’t be able to understand what you’re looking at. But if you pick an appropriate attribution model that fits your business, budget and resources, then you will find yourself with data that let’s you refine your marketing plan in very useful and interesting ways!
We hope you found these points, and the slides below, useful, and wish you luck with your campaigns!
We created Royku to train marketers in data-driven marketing.