Comparing attribution of multi-channel models

Comparing attribution of multi-channel models

In this article we examine the reasons of non-accurate assessments of marketing channels using multi-channel attribution models in GA (Google Analytics) and create more dependable algorithms. This can be of use to marketers and analysts looking to come up with independent models of evaluation for marketing channels.

Last Click is the most famous evaluation model for marketing channels of the last touch point (Last Click) or the last indirect contact (Last Non-Direct Click) that GA uses by default. While it may appear only logical to assign 100% value to the channel, that was last before the transaction.

Although this will not address important queries like:

Why lowering the budget of Channel A has led to a fall in the revenue of Channel B?

Why raising the budget of Channel B has led to a rise of CPA of Channel C?

The obvious answer being channels have an influence over one another with their actual contribution is different from the attribution of the Last click Channel.

What is causing this?

Let’s examine the Path Length. In all likelihood, the report has a similar pattern with more than half generated income from the orders committed after the purchaser visits a website a number of times. That shows that a portion of the visits in the particular order that led to a sale are not estimated as the last source gets the full value. To evaluate the problem’s extent it is essential to calculate the session’s proportions that are not evaluated through the Last Click Model. There is no need to try and figure out why incorrect decisions are made based on inaccurate channel values when the Last Click model of attribution is used.

Besides Last Click the user has made to come to your site, GA can find out the impact of all channels via Assisted Conversions. He or she makes multiply actions with your site before converts. Using GA a report can be generated on Assisted Conversions to verify the income derived from the channel, despite it not being last in the chain. But certain channels can have doubled impact in the report, easily overstating the impact of certain channels. GA has no answers for these queries: it would be a challenge to convince your superiors in the accuracy of the report.

For those marketing specialists that are of the opinion that Last Click model of attribution is not adequate, the good news is GA offers various other Non Last-Click Attribution Models as well.

Since the CFO would not be happy it is best to opt for a single model for all channels. However, Position Based and First Interaction models have similar disadvantages to Last Click as they do not take into account the input of most of the sessions that have an influence on sales funnel process when the customer is on the way to placing an order.

GA 360 offers an attribution model that has the objective to resolve the limitations of the models mentioned above with a Data Driven Model. This is a more progressive and balanced way than all other previously mentioned methods. While it seems just ideal, it is fine as long as there does not arise a need to explain why the Data Driven Model displays an accurate value and not any other. Also why Google servers count it, and the one you cannot verify, should be taken into account as the most accurate.