Disruptive Marketing Blog

Attribution Models for Advanced Users

What are attribution models?

The answer is simple and crystal clear, as you might imagine. It’s knowing how one thing influences another thing in order to achieve an objective. For example:

• How much influence does TV have on Google searches when it comes time to obtain a business contact?
• How much influence does a banner campaign using the best media outlets of a country have on brand campaigns in search engines?
• How much influence does social media have on all traffic sources in order to obtain sales?

Attribution models are problematic because they only measure clicks, not intentions. In other words, if a user clicks on one traffic source, and then on another, it’s taken for granted that the first click has influenced the second click. It might be, however, that the user has merely browsed through that first click. Is there too much importance being given to one click?

 

Customer Journey

As we have already mentioned, if an attribution model is the study and analysis of the influence of some channels on others in order to achieve an objective (conversion), and we analyze this concept at the level of the individual customer, we are referring to the Customer Journey, in other words, where and how many clicks are needed for a user to actually buy something.

Because of that, it is just as important to measure which channels a user has used when he/she actually buys something, as when they do not.

 

How are conversions tabulated?

The most common and standard method is to assign conversions (lead, sales, etc.) to the last click, since it is the click closest in time to the objective.

But, if we follow this criteria, we find ourselves falling into a very common error: conversion duplication.

Let’s imagine that we’re working with Google Analytics in order to measure web traffic. Likewise, we’re using SEM as well as email marketing campaigns in order to obtain sales. A user takes the following “journey” and winds up converting:

 

77

 

These are the statistical results:

 

777

 

Now we begin to play a numbers game. We start off with the logical idea that every traffic acquisition channel (that uses a statistical system), will always want to take the credit, because when its traffic results in a sale, regardless of whether the sale stems from the last click, the acquisition channel will assign the sale to itself. So it follows that:

• Google Analytics will count the sale for the last channel (click).
• Payment Search statistics will tell us that the sale belongs to them.
• Electronic Mail will give themselves the nod.

If we add the statistics we get from the platforms that we are using, we see that we have sold 1 + 1 = 2 sales, when we have really only obtained one sale. Therefore, it is of primary importance to measure and make decisions about a system that has a global vision of at least our payment traffic.

Now, let’s imagine an Advertiser that is running campaigns in Adwords and on an affiliate platform. How many of our sales could actually be duplicate sales? Or even worse, if we measure post-view banner conversions of a banner on a major communication media outlet, and we confirm that there is an overlap with traffic from an Adwords campaign, we will note that a high percentage of sales are shared. However, it might be possible that this banner has not influenced the user during the buying process, because the user hasn’t even seen it. But, it has been shown to an audience so large that it influences every other media outlet.

How do we detect then,at a glance, the need to have an attribution model system in place?

A key point is to know whether we are working with investments that have an impact on our bottom line from two or more traffic acquisition channels. SEM + Affiliates, SEM + Email… But, if you really want to base your decisions on numbers, I advise you to access the Google Analytics report: Conversions > Multi-channel Funnels > Interaction Paths:

 

7777

 

Out of a total of 19,722 conversions, 50% of them require more than one click. Therefore, it is likely that if we are working with a variety of traffic acquisition channels, the user has browsed through those same channels. So, it would be interesting to take a serious look at the subject of attribution models.

But, without diminishing what we have already said, we shouldn’t lose sight of the fact that, although 50% of conversions needed more than one click for conversion, the other 50% only did need one click for conversion. Therefore, not only do we have to analyze our attribution models, but we also have to analyze our CRO so that these one-click conversions improve our success rate month after month. A basic example in order to see how my conversions are distributed over different channels is to access our Google Analytics and review the report:

Conversions > Attribution > Model Comparison Tool:

 

77777
Let’s do a comparison of an attribution model based on assigning the sale to the last click as opposed to a lineal one. For example, if we distribute the bulk of conversions equally between all participating channels, we see how the e-commerce channel would have 13% less conversions and payment traffic would have 12.92% more of them. Isn’t it great to have a 12% increase in conversions by merely changing a button?