Attribution Modelling – How Different Models Work

A common question organizations using various online marketing channels have is, “which one works best for my business?”

“Most of my traffic comes from paid search, they convert less. Conversion from the traffic through SEO efforts is higher. So, Do I have to invest heavily in paid search or SEO?”

Almost all businesses face similar questions on choosing the right marketing mix for their business. Digital marketing opened up various new avenues providing various touch points with the customers. It also becomes important for the marketers to choose the right mix of all the marketing channels to provide higher ROI for their marketing spend. This is where attribution modelling helps the businesses.

Attributions modelling is the science of understanding the value of each customer touch point which leads to conversion. Google Analytics allows you to use various attribution models to analyze the impact of your various marketing channels. Here is a list of attribution models defined in google analytics

Last Click Attribution Model

Last click attribution is one of the most widely used attribution model. It assigns 100 percent of the conversions achieved to the last customer touch point. It does not consider various other touch points the customer has followed during the conversion path. Though it is widely used, it is one of the most inaccurate methods to attribute success to your various marketing channels.

Last Non – Direct Click

Last non – direct click model assigns the credit of conversion to the last touch point before the conversion which is not direct traffic. For example, if a visitor visits your site through a paid ad, social campaign and then visited your website directly, while the last click attribution model gives 100 percent credit to direct channel, last non – direct will give the complete credit to social campaign. Instead of visiting the website directly, had the user done that using a display ad, both last click and last non – click give the complete credit to display ad.

Last non – direct click is defined based on the assumption that, if a visitor was already influenced to make a purchase decision if he/she is visiting the website though direct route. So, the last non – direct click assigns the credit of conversion to the channel which has led to direct visit by the visitor.

Last Adwords Click

This model can be used to understand the effectiveness of the AdWords campaigns. This model gives all the credit of a conversion to the last AdWords click prior to the conversion.

First Interaction

First Interaction model as the name suggests, assigns all the credit of the conversion to the first customer touch point in the conversion path. However, it does not really explain the whole story behind the conversion. One of the major criticism such model faces is, if the whole credit of conversion is given to the first interaction, the conversion should also have happened during the first interaction. However, visitors go through various conversion paths and touch points before the conversion.

Linear

While all the models defined till now try to attribute the conversion credit to a single customer interaction in the conversion path, linear model gives every interaction of customer equal weightage. For example, if a customer is exposed to paid search, social channel, display ads and direct search, each channel is attributed a 25% credit for the conversion. So, each interaction is considered as important as the others in the conversion path.

Time Decay

Time decay model is one of the more advanced methods of attribution modelling. This model is defined based on the assumption that the closer a customer touch point to the conversion, the more influence it had on the conversion. Based on the timing of the touch points in the conversion path, the algorithm assigns the credit in a decreasing order starting from the last interaction to the first.

You can even customize this model by setting half – life decay age which gives half the credit of the conversion day click to the interaction that happened prior to the conversion during the specified period. You can also specify a look back account to define the number of days prior to conversion that are to be considered by the model. You can also apply weightage to the interactions based on the user engagement parameters like time spent on website and page depth.

Position Based

Positional based model is a hybrid model of last interaction, first interaction and linear model. The model assigns 40% of the credit to the last interaction model and first interaction model and remaining 20% is assigned equally among the remaining transactions. The closure cycle generally follows the process of creating an awareness, after which the customer keeps looking for more information about the product/service which leads to final conversion. The position based method assigns more credit to the channels which created awareness (first interaction) and the touch point which lead to conversion (last interaction).

You can also customize the model by specifying the credit you want to assign to first interaction, last interaction and middle interactions. You can also specify the look back period and adjust credit based on the user engagement.

While google analytics does provide a variety of pre-defined attribution models, it is important that you analyze your business goals and KPIs before implementing the best model for your business model. While last interaction and first interaction almost always provide skewed results, time decay and position based models can be customized to define attribution models which can address your needs better.

Related Post

Author: Mahesh Tadepalli

Mahesh Tadepalli brings the analytical edge to the marketing campaigns we manage. A holder of Master’s Degree in Mathematics, he plays a major role in harnessing the marketing data for effective decision-making. His experience in working with tech startups and understanding of marketing have proven to be instrumental in defining marketing technology solutions for companies across various industries.