One of my favourite conversations in digital is about attribution. Leading a cross channel team, allows me to see first hand, how channels can argue over who contributed more than the other for that sacred conversion. Once a conversion happens, everyone wants the credit for it: offline, search, display, remarketing, email etc… The question is – is there 1 right answer to which channel deserves the credit?
And the answer is – there is no perfect answer.
Many companies will come and try to convince us that their attribution solution is the silver bullet. Gold dust sprinkled on top of it. A secret magic ingredient better than anyone else. No matter what, the one thing I have learnt through the years is that perfect attribution is impossible. It is just a matter of establishing an actionable model, you as a marketeer feel comfortable with. A model, which will inform future budget allocation and allow for a test and learn environment. Every model has its positives and negatives – established brands, new comers, custom modelling. There is a lot of choice out there and prices vary significantly based on what you choose.
Attribution is all about allocating credit or a weight to each of the channels. The challenge comes in calculating the exact (accurate) credit each channel really deserves, and adding the concept of ‘time’ to the model. There is also another challenge, which I think is the most signficant one yet – external factors.
I remember when I was at Uni and I did my credits in Economics, and Economic modelling etc.. there was one phrase I was fascinated by: Ceteris paribus. Ceteris paribus is a Latin phrase meaning “all other things being equal”. Attribution is very similar to an economic model – you can primarily look at 1 factor, assuming everything else is held equal and does not change. However, as in an economic situation, in real life (marketing) nothing is held constant while you test your campaign 🙂 Things constantly move and one cannot be aware of everything at all times. So, as usual some assumptions need to be made.
Let’s say a travel campaign is launched cross channel and platform. It goes really well and then the results decline. As a marketeer, you can assume that the campaign is not working, the channel mix is not correct, the creative is tired. Your attribution model can show if any of the channels are seemingly not pulling their weight.
However, what you will never see is that it has been sunny the past week and people are just not in the mood to sit on their mobiles/PCs, as they are busy having fun outdoors. Will everyone know this – no. Will your attribution model know this – no. Obviously this is a simple (very simple) correlation example.
Connecting online to offline is a much more common struggle for digital marketing at the moment. How many people saw this banner? How many called the none dedicated number after seeing it? Did the person taking the call, allocate the booking code properly so it can at least be allocated to the right channel grouping? The list can go on and on. The moment online needs to connect to offline, human error becomes an issue. Humans cannot be controlled the way you can control an algorithm or a model.
To conclude – attribution is a little bit of a unicorn. However you can create your own version of it – a good breed white horse and attach a horn, sprinkle some glitter and off you go. A perfect attribution model has been created and it is all yours. How much you pay for that, ah well that is up to you 🙂