I recently read an interesting white paper about building an attribution model that works. The paper got me thinking about attribution modeling, and why it is such an important asset for any company (both off and online). As such I thought I would give a brief introduction to attribution modeling and give three tips to help your company build an attribution model that works. To understand attribution modelling you need to go through a two step process. The first step is to look at the problem it solves, and the second step is to have a description of what attribution modeling is. The best example of the problem that attribution modeling aims to solve is found here in a blog post by Eric Krell. He describes this scenario: "Imagine a simple Internet purchase: a $100 transaction from Zappos.com. A reader on The New York Times website sees a display banner for Zappos. It makes her think about sandals. She surfs on Zappos and other sites. After a beach weekend, she decides to order the sandals she looked at by typing “Zappos” into Google, and makes a purchase. One hundred percent of this advertising credit goes to the paid brand search ad, which was simply the endpoint before checkout, not the starting point: the ad that first stimulated her purchase. The marketing executive running the Zappos’ Internet advertising budget gets a false read, with misleading data on which advertising expenditure truly drove revenue." As we can see the problem in this scenario is the fictional executive at Zappos gives all the credit to the last place the consumer receives and acts upon the message (not surprisingly called "last touch attribution"). The fictional executive completely disregards the multiple touch-points that the consumer encounters before, and because of this has faulty information he provides marketing strategies that are not fully optimized to the real world . "Oh no" you say- how do we solve this problem? Well have no fear because attribution modeling is here to the rescue! Attribution modeling looks to take you away from using last click attribution and to look at the multiple channels and the number of times your customer encounters with your message. By carefully examining data, weighing the different touch-points (including the time between the message and the purchase) and comparing these factors you begin to build an attribution model! Attribution modeling hopes to shed light on questions such as: 1. How much credit do we give to each of the messages before the final purchase? 2. How does time factor in? - was the first message more potent than the others? 3. How do we track these messages? The answers to these questions depend on your data set, the channels you use, the customers you are trying to reach as well as a variety of other factors. However, we can give you 3 pieces advice that anyone can use at any company to begin building an attribution model that works: 1. Don't only rely on on the last click to give credit for the purchase. This is the basis for building an attribution model and seems the most obvious to anyone who read the example. However, many companies never even think about giving credit to more than one source. Even Google Analytics is set up for last touch attribution! To learn how to reconfigure GA check out this article from SEOmoz (warning - it gets pretty technical). As we saw in the example, the executive at Zappos gets a false read if he only gives credit to the last click that occurred before the purchase. The idea is to switch your mindset and realize that many factors effect the consumer's decision to buy. These factors range from the different capabilities of the channels, to the time between receiving the message and the purchase, to the message itself. If you begin by seeing the whole picture then you have a head start. 2. Track, track, and track some more. Your company may not be set up for attribution modeling currently, but when it is you will need a decent size data set to begin. We recommend tracking everything that you can and tracking it all with one tool. Since different analytic tools can offer variations on the same data it would be a disservice not to track the users "apples to apples". This way when the company is ready to look at the different channels, all the information will be right there and will be uniform. Track as much as you can as often as you can, and you will begin to see patterns that you can compare. These comparisons are the seeds of building an attribution model that works. 3. Know what your data means. Tracking the data is a fantastic start, but to build an attribution model and optimize it you need to know what the data means. There is a lot of work to truly understand data (believe me, at DO we know!). Understanding the nuances in the data and the touch-points the customer interacts with takes can take years of training, and is quite time consuming. There are a variety of great resources to learn about data analysis and I personally recommend going through the official Google videos - they are informative and presented in an excellent way, or read Avinash Kaushisks' blog . However, if you don't have the time or inclination to spend pouring over analytics we have a team of experts here at DO that would be happy to help. Feel free to contact us to help your with data analysis and all of your digital strategy needs.
More from the
Why TACoS is an Essential Metric for your Amazon Toolbox
Amazon / February 19, 2020View Blog Post
Driving Digital Marketing ROI with Conversion Rate Optimizat...
Strategy & Planning / February 13, 2020View Blog Post
Partner Q&A with Teikametrics – A Retail Optimiza...
Amazon / February 7, 2020View Blog Post
Extend holiday sales spikes into January 2020 with these tip...
Ecommerce / November 14, 2019View Blog Post