Understanding the Analytics Attribution modeling

2018/03/27 07:54

attribution modeling

The number of channels customers are using to connect with brands had significantly increased in the last couple of years. We now have blogs, social media, search… and the list is much longer. With so many customer points of contact, there are also more questions about the real value of the dollars spent on each medium. Several years ago, if you wanted to measure the effectiveness of your campaign all you had to do was:

  • Measure the traffic and sales before running the campaign
  • Run the campaign
  • Measure again the traffic and sales
  • Assess the differences

On the other hand, today it is much more complicated to get meaningful insight on how customers are led to conversion, because of the increasing number of channels.

But how can I do it?

Attribution analysis can provide you the answer you are looking for. It brings together data from many sources and it assigns credit to the channel or channels which led to conversions. Its main objective is to help you understand the behavior of your visitors and establish the most powerful marketing channel to invest in. With the multitude of channels at hand today and ubiquitous availability of information that a prospective buyer has to digest before making a purchase, it is likely that there is no single channel that should be credited for the conversion. More likely, there are several channels that contribute indirectly over time to the final result and the credit for the conversion should be distributed between them.

Why is Attribution Analysis Important?

Let’s consider a simple example where, supposedly, you would like to buy a dishwasher for your new home.

The first step… you will go to your best friend Google for advice and perform a search. There will be lots of search results, including Homedepot, Lowes, Bestbuy and the like. You click on one of the organic listings and go to Homedepot website. While browsing through the available models, your boss calls you and asks for the latest sales report. Next day, you start again to search on Google, but this time you click on one of the PPC ads and land on the Homedepot page. Since all the models you like are either over your budget or out of stock, you decide to subscribe to their newsletter and wait for a discounted price or the arrival of an affordable model. After two or three days you get an email newsletter from Homedepot announcing a 25% off. You decide that this is the moment and click on the “Buy Now” button. This redirects you to their webpage where you make the purchase.

The point I’m making here is that, even for a simple purchase, the buying process is more complicated than “I know the exact product and will buy it from this retailer”. A prospective buyer will click on multiple ads before being converted. This makes it hard to properly allocate credit to all the marketing channels involved.

Attribution Analysis on Google Ads

The vast majority of marketers measure the success of their marketing campaign using only “last click” information, giving all the credit for the conversion for the very last clicked ad. Anyhow, this approach ignores all the steps that the customer has passed through before being converted. As shown above, there might be quite a few, even for a simple purchase.

Fortunately, there are tools available to help you do that. Today we will have a look at how to perform attribution analysis on Google Ads using attribution models.

What is attribution model?

Google AdWords Attribution model is a set of rules that provides granularity on how the credit for conversions should be attributed.

The following attribution models are currently available to assist you to understand how prospective customers search for your products:

  • Last click: all credit is given to the last clicked ad and corresponding keyword. It is the easiest model to apply but it ignores important pieces of information.
  • First click: the exact opposite of the last click, all the credit goes to first clicked the ad and corresponding keyword.
  • Linear: credit is equally assigned to all clicks.
  • Time decay: provides a weighted distribution with a 7-day half-life, assigning more credit to clicks that happened closer to conversion. It is useful for businesses with longer conversion cycles.
  • Position based: 40% of the credit goes to first and last click and the remaining 20% is spread across the other clicks.
  • Data-driven: Back in May 2016 Google introduced machine learning technology named Data-Driven Attribution (DDA) which helps you calculate how much credit should be assigned to each click before a conversion. Instead of crediting for conversion only the last clicked ad, DDA uses data from your account to find out exactly which adds, campaigns and keywords make the greatest impact. By comparing the paths chosen by customers which end up converted to the ones who don’t convert, DDA identifies the pattern or patterns that lead to conversion. This attribution model only works for accounts with sufficient data.


As I showed above, the last click is the simplest and also the most limited AdWords attribution model you can use. To assist you in reaching your clients earlier, Google Ads provides not only several other attribution models but also bring the means to compare the results of two attribution models. This can be achieved using attribution modeling report in your AdWords account.

For example, one of the simplest and most useful comparisons you can make using attribution modeling report is a side-by-side comparison between the last click and linear attribution models for the same campaign. You will gain useful information on the keywords and ads that guided the customers through their path to conversion.

When starting with attribution analysis it is essential to clearly define the success factors. Keep them simple, quantifiable and work continuously to reach them. Attribution analysis is not a one-time effort, it is something that you must do all the time and, with each iteration, get closer to your success factors.