In an earlier blog Which Attribution Model Should I Use, we gave an overview of the more popular single-touch and multi-touch marketing attribution models. In this blog, we discuss about three common attribution models, and look at the pros and cons of each one.
Based on many years of experience running e-commerce stores, we have also added unique features that make each of the models we support more valuable to you as a marketer.
First Touch Attribution Model
The First Touch (First Click) attribution model attributes a conversion (e.g., a purchase) to the first interaction made by each user.
In this diagram of a customer journey, the First Touch is Organic Search and it gets credit for the purchase:
Pros:
The First Touch model allows the marketer to see which channels or campaigns perform best in bringing new purchasers to the site.
In short, First Touch provides guidance on where a marketer should invest to best attract new users.
Cons:
Typically a small portion of your purchases will come directly from a first click. As in the diagram above, there are most often a series of clicks that lead to a purchase. But in the First Touch model, the first click takes 100% of the credit. This ignores the other interactions that led to the sale (distributing credit across each participating touchpoint is left to multi-touch models).
Unique Value-added Features:
Since the First Touch model highlights the value of new purchasers on a site, understanding their behavior after the first purchase is critical to assessing their true value.
To understand their subsequent behavior and their true value, our Dimensions Attribution software platform tracks their subsequent purchases and credits these back to the channel or campaign that was credited with the initial purchase.
In the above diagram, First Touch attributed the purchase to Organic Search. This customer’s subsequent purchase (illustrated below), would also be attributed to Organic Search.
This allows the marketer to get the true value of purchasers based on the revenue they generate, and in the case of paid campaigns the customer acquisition cost, average order value, and return on ad spend (ROAS). It also helps marketers determine which channels are more successful at building/recruiting a strong customer base.
To complete the picture, as well as differentiating the revenue from initial and subsequent purchases in its First Touch model, Dimensions breaks down the number of purchasers and orders associated with both first and subsequent purchase revenue.
Last Touch Attribution Model
The Last Touch (Last Click) model will attribute a conversion to this last interaction immediately Prior to a purchase.
In the first customer journey diagram we illustrated above, the Last Touch prior to the purchase is Paid Search, so this channel gets credit for the conversion:
Pros:
The Last Touch model allows a marketer to see which channels or campaigns perform best at converting a user to a purchase, whether for initial purchases or subsequent ones.
Last Touch helps to guide the marketer where it is best to invest to potentially increase the number of purchases on an e-commerce site.
Cons:
The Last Touch Model sees 100% of the credit for a purchase going to the last interaction preceding either an initial or subsequent purchase. So Last Touch ignores the other touchpoints in the user journey that led to a user converting, including the First Touch that drew the customer to your site/product in the first place.
Unique Value-added Features:
The Last Touch model always gives credit to the touchpoint that immediately precedes a purchase, regardless of whether the touchpoint is for an initial or subsequent purchase.
In the following subsequent purchase, Last Touch would attribute the purchase to Email:
Differentiating initial and subsequent revenue from Last Touch conversions presents a more detailed picture of the touchpoints that perform well in directly converting customers.
Again, in Dimensions’ Last Touch model, customer acquisition cost, average order value, and return on ad spend (ROAS) are broken down, along with the number of purchasers and orders associated with both first and subsequent purchase revenue.
Dimensions also offers a Last Non-Direct Touch model. This model ignores Direct touchpoints when identifying the Last Touch interaction except in the case where it is the only one. This model allows the marketer to focus on the Last Touch interactions they can directly influence, as building Direct traffic is a longer term process largely based around building consumer loyalty and brand identity.
Machine Learning Attribution Model
In our earlier blog Which Attribution Model Should I Use, we reviewed some of the multi-touch models, such as the Linear model which gives equal credit to all touchpoints.
But these models are “heuristic”, simplified “best guess” or “one size fits all” types of approaches to creating multi-touch attribution models.
Building a custom model uniquely reflecting what is happening on each store’s site is time-consuming to build and costly to maintain.
Machine Learning (ML), a form of Artificial Intelligence, solves these issues. Machine Learning facilitates building an attribution platform that can more accurately predict outcomes using historical data (in our case the sum of all user journeys). Using these inputs ML can then generate a weighting for each channel or campaign and predict future performance of the channels and campaigns deployed by each site.
Pros:
Dimensions ML model gives the marketer a multi-touch model that is based on user behavior on each store’s site. Because all touchpoints are not equal, ML can identify the touchpoints that are more important in a user’s journey. Using this analysis, ML can then build a model that weights each channel and campaign based on its unique characteristics.
Because ML is designed to handle large volumes of data, the models that are built for each site can be generated on a daily basis.
ML thus provides an accurate multi-touch model, day by day, site by site. And attributes credit across touchpoints to channels and campaigns that reflect their contributions to a purchase.
Cons:
None, Dimensions ML is the gold standard.
Unique Value-added Features:
Again, the presentation of data (initial and subsequent revenue; customer acquisition cost, average order value, and return on ad spend, ROAS) is consistent across the Dimensions platform.
Dimensions’ ML implementation is a game changer because of its unique ability to build future revenue projections for each channel. These projections are built by determining the probability of purchase for a site’s past visitors (purchasers and non-purchasers) based on their behavior on a site, and applying these probabilities to future visitors.
Summary
These are the three models our customers tend to use when optimizing the channels and campaigns that are working, adjusting the ones that aren’t, and leveraging opportunities to improve the return on their marketing investments. To learn more about other attribution models, read Attribution Models Compared.
