Croud’s Analytics Manager and Data Head Igor Korovenkov, discusses ways of transforming Google Analytics Multi-channel funnels information into actionable data.
I must confess: I have been a bad, bad, data analyst and should be punished by being scraped out of my number-filled comfort environment and put into the sales department to answer phones, cold-call and work with… gulp… people. Unpredictable, irrational people… My sin? Well, ONE of 397,666 sins – using last touch attribution to determine digital media value. And boy, did I use it shamelessly.
The penance for my sins came after Google added Multi-channel funnels to the arsenal of Google Analytics (GA) tools and after I finally learned to extract use from it. Well, I am glad I did it when I did, because, you guys, Multi-channel funnels section is extremely useful for anyone who wants to deliver greater value with analytics reports, but does not show enough dedication and tender love for data accuracy to shell out £90k annually for GA’s premium package to be saved from the hot pits of sal… I mean, hell.
In work, I am always torn between delivering accurate numbers and generalising to valid, representative abstractions of a largest possible scale. To put it simply, I aim to describe precisely in numbers a real life entity. Simple, right? Digital media value is exactly this kind of entity. Determining it always triggered the side of me that aimed for generalisations, because I could never let go of the thought that last touch was only one way of looking at attribution and digital media value.
With Multi-channel funnels tools, you can zoom out a notch and get a glimpse of the bigger picture, since they allow to look at first click, last click and assistance scenarios separately. For a quick reference, first click scenario attributes a transaction and its value to the medium that was the first interaction in a chain, that led to a purchase. By contrast, last click scenario attributes conversions to the medium that preceded a transaction. Assistance scenario regards all the interactions in the chain that served as an intermediary – they did not lead to a transaction, but passed a visitor further on.
So for example, in a three-step (medium A) -> (medium B) -> (medium C) interaction chain that resulted in a transaction, medium A would be attributed the transaction in the first click scenario and medium C – in the last click scenario. Medium B would be attributed the transaction only in the assistance scenario. As these are three separate scenarios, GA attributes the whole value of a transaction to a medium in every case.
If you are as obsessive about profiling things and generalisation, you, as I did, would wonder if there was a way to aggregate the data across different scenarios to get a “more general” digital media picture. A more simple and straightforward way in this case would be to simply sum up the numbers and enjoy the result. But, after giving it some thought, I figured that it would make much more sense to assign different weights to the scenarios. After all, being proper entrepreneurs that we are, we want conversions, and not just interactions! Another thing is that different digital media perform differently by default, and we choose them depending on the goals we want to achieve, so assigning equal values would mean dismissing this fact. Lastly, it would be simply arithmetically correct to treat the final picture as 100% and make those three scenarios comprise this number.
My current formula is 40% – 20% – 40% for first click, assistance and last click scenario respectively. It allows me to aggregate digital channel activity data (expressed in the number of transactions) and calculate values of digital media for the business, expressed by the value of transactions that they drove. Any shifts in the digital media management will reflect on the numbers, and we shall be able to notice them and analyse. We could even calculate average transaction values on every medium and track how they change!
Before you get to grumbling about possible limitations of this approach, think that it pushes back the day you succumb to the temptation of spending bajillions of money for GA premium and allows you to go beyond the last click attribution. I would feel better about that alone!
Although I do not like my dreams and hopes shattered, please feel free to point out any (oh, how bold a word…) caveats in my approach, criticise… or praise!