PMG Digital Made for Humans

Finding the Best Attribution Model

8 MINUTE READ | January 17, 2018

Finding the Best Attribution Model

Author's headshot

Eric Brown

Eric Brown has written this article. More details coming soon.

Half the money I spend on advertising is wasted; the trouble is I don’t know which half.__-John Wanamaker

The modern shopping process is extremely different than the shopping process used just 20 years ago.  Back then, the Spice Girls topped the charts, Beanie Babies were all the rage and you were likely to save your computer files on a floppy disc. As computing changed, e-commerce changed with it. Now, Google crawls billions of pages a day, music fans save their songs to the cloud and libraries of information are available on a mobile user’s smartphone.

Shoppers can now access product details, reviews, and pricing information at a moment’s notice.  Innovations such as these have changed the way people shop with many users accessing e-commerce sites numerous times before making a purchase decision.  Many marketing and advertising analysts have been a bit slow to capitalize on this shift in behavior and should consider a change in tactics.

An attribution model is a rule, or set of rules, that determines how credit for sales and conversions is assigned to touchpoints in conversion paths.  Attribution models help an analyst decide which touchpoints (in this case, traffic sources) are responsible for conversions on a website.  This data can then be used to make decisions such as which traffic sources or campaigns should receive greater or less spend.  Simply stated, the model helps in optimizing marketing spend.

Google Analytics is the tool of choice for many marketing and advertising analysts.  Google offers a useful suite of marketing reports for analysts to use out of the box.  Most use Google’s Acquisition reports which use what is called a “Last Click” Attribution Model.  This model awards all credit for an action to the source directly attributable for that action.  In this case, an “action” could be considered a conversion, an event, a successful session; anything that is considered a success metric.

This is problematic because the user might have had multiple interactions or exposure to media with a brand prior to this “last” action.  In many cases, these previous interactions influence the user’s decision to convert and therefore should be given at least part of the credit for the eventual conversion.  In other words, entire path of a converting user should be considered, not just the last interaction.

Unlike Last Click Attribution, First Click Attribution awards all credit to the first interaction a user had previous to a conversion.

The First Click Model differs from the Last Click Model as only the first click in the “look back” window would get credit for the user’s eventual action.  The “Look Back Window” refers to the number of days prior to the conversion to analyze interactions for the user’s eventual action.

By default, Google Analytics will limit the amount of time to “look back” to 30 days (however, this can be set as far back as 90 days for analysts using Google Analytics 360 Suite).  A 30 day or shorter lookback window is sufficient for most sites especially for sites that offer products with a short consideration windows like low priced electronics, apparel, and low-cost retail, however, sites that offer products with longer consideration windows like high priced electronics, furniture and B2B should use longer lookback windows.

While the First Click and Last Click Attribution Models are good for finding the interaction that started a consideration cycle or the session where a conversion was made, it is safe to assume that each interaction in the consideration cycle will have some influence on the purchase decision.  Because of this, it is also safe to assume that each interaction should claim a share of the credit for the eventual conversion.  Google Analytics has 3 Multi-Channel Attribution Models for use by default.

Linear model give sequel credit to all interactions in the Lookback window

Time Decay model gradually gives more credit to interactions closer to the conversion session

Position Based model gives the highest credit to the first interaction and the conversion session


Analysts with access to Google Analytics 360 Suite can use the Data Driven model which uses a proprietary algorithm to find the best model for the traffic data applied to it.

Like many issues analysts will face, there is no “one size fits all” attribution model.  Every site is different and every marketing team faces their own challenges.  There are marketing problems that the aforementioned models will not answer for.  Because of this, Google Analytics provides a tool for creating custom attribution models.

There are a number of options available for tweaking the model to fit an analyst’s need.  This comes in extremely handy when dealing with ad impressions, which have proven to be hard to analyze prior to Google Analytics Attribution reports.

Google Analytics includes reporting tools for this type of analysis in the Multi-Channel Funnels and Attribution reporting found under the “Conversions” tab in Google Analytics.  The Model Comparison Tool report can even be used to compare different models (i.e. “last click”, “first click”, etc.).  Cost data can also be added to most of the Attribution reports.  This allows an analyst to know not just how actions and revenue are attributed, but also how ROI is attributed.  Cost data can either be manually uploaded to Google Analytics or can be automatically brought in via Google’s integrations with Google Adwords and Google Doubleclick.

With all of the visibility that Google provides for these attribution models, there is no silver bullet or all sizes fit all model.  An analyst’s goals should be put into consideration when choosing a model.  Here are some common goals for marketing/advertising analysts.

Goal 1 – Acquisition and Brand Building

When acquiring new users, exposing the brand to a new audience should be the main focus.  These types of campaigns are usually poorly attributed since these campaigns usually bring in users that do not have much history with the brand.  While these users are usually not highly qualified and therefore lead to low session conversion initially, they are new users that are not likely to know about the brand and have the possibility to be highly qualified in subsequent sessions.

In this case, an analyst would want to choose a model that awarded more credit for the initial interactions of the user’s consideration cycle.  First Click and Position Based would be good candidates in this case.

Some analysts will have to put a considerable amount of work into making sure their users continue visiting their site multiple times.  This is especially important for content sites, B2B sites, sites that showcasing high priced items such as cars and real estate listings.  This type of decision is not impulsive, so the user will require a relatively long consideration cycle paired with a significant amount of research.  The Time Decay model could be used, but the Linear model is probably best as it assigns equal credit to interactions in the middle of the consideration cycle as the first and last.

If an analyst’s goal is to increase the number of sessions from users that are already aware of the brand, more credit should be placed on interactions that occur later in the consideration cycle.  These users are already aware of the brand, are highly qualified and could have possibly already converted on the site on a prior session.  In this case, the Time Decay model should be considered as it places higher credit to interactions late the cycle and also gives credit to interactions immediately prior to the last action.

As previously mentioned, there is no “one size fits all” attribution model.  An attribution model should be selected with a marketing/advertising goal in mind.  With that said, it’s up to each individual analyst to find the correct model for their needs.  While he/she can assume that acquisition campaigns are probably best measured by models that assign more credit to interactions that come early in a user’s consideration cycle, he/she should test to make sure their assumption is true.  There some major reasons why testing is the best method for finding the best attribution model.

Stay in touch

Bringing news to you

Subscribe to our newsletter

By clicking and subscribing, you agree to our Terms of Service and Privacy Policy

  1. To prove effectiveness of attribution in general

    • Attribution models are intended to help the analyst ensure that spend is being allocated as best it can be.  Because of this, proving effectiveness is paramount.  Tests should always be run to ensure that there is a difference between attribution models in comparison to last click attribution and that said difference is positive.

  2. To prove effectiveness of one attribution model to another

    • Just like proving the difference between other attribution models to last click attribution, tests should be run to prove the difference between each model.  If there is no difference between each attribution model, it could be that either the “look back” window is too short to analyze multiple interactions or that users mostly use a single channel to interact with the brand.

  3. To customize models for an analyst’s needs

    • With Google Analytics providing custom attribution models, an analyst can use a number of options to create the model to fit his/her needs.  Vigorous testing should be applied in this case to find the settings that will ensure success.

Related Content

thumbnail image

Why You Should Learn SQL

4 MINUTES READ | July 5, 2019

thumbnail image

3 Tips for Showing Value in the Tech You Build

5 MINUTES READ | April 24, 2019

thumbnail image

A Beginner’s Experience with Terraform

4 MINUTES READ | December 20, 2018

thumbnail image

Tips for Holiday Reporting Preparedness

3 MINUTES READ | November 5, 2018

thumbnail image

Why Testing Matters

4 MINUTES READ | October 31, 2018

thumbnail image

Navigating the Amazon Ecosystem

2 MINUTES READ | September 10, 2018

thumbnail image

Our Approach to Marketing Automation

7 MINUTES READ | November 16, 2017