How a 19th century Russian helped solve your marketing attribution problem
Updated: Jun 1
The American department store magnate, John Wannamaker, famously bemoaned: “I know half my marketing budget is wasted. I just don’t know which half.”
Almost a century after his death, marketers still struggle to assign true value, or weighting, to their various marketing activities and the customer interactions they support. Many attempts have been made to identify the significance of such interactions. Whilst they all claim to have merit, these are often ‘click’ based methodologies with little or no statistical validity whatsoever. They simply don’t consider what influence, interaction and relationship each channel contributes towards a purchase.
Importantly, they don’t account for when a customer researches online then purchases in store or contacts your call centre.
Markov chains - a stronger attributional link
At UniFida, our approach to weighting for marketing attribution is to use a technique developed by the 19th Century Russian mathematician Andrey Markov, known as Markov chains. This methodology is widely known and respected and has found applications in everything from biological research to economic forecasting.
We use Markov chains to analyse sequences of events across hundreds of customer purchases to determine the impact each event has on the purchases they lead to.
We normally take the 90-day window before each order and include every recorded online or offline event. The chain can include browsing activities, email opens, retail visits, SMS messaging, click throughs from social media, and anything else we can stitch together.
Knowing the impact of all the events of a particular type, email opens for example, lets us assess very accurately the impact of the overall email channel, when used in combination with the other channels and media we are using. The value of the orders delivered by email, when set against the cost of email, gives us the ROI for that particular channel.
We repeat this approach for all online and offline channels to get to an understanding of the ROI delivered by each of them.
Addressing the what ifs….?
What if someone opens an email several times? Does this count as multiple events or just one? We normally regard multiple opens in a short space of time as a single event.
Another dilemma presents itself when the customer goes straight to the website of the supplier. If this isn’t preceded by other events then we give credit to the website and the overall strength of the brand. On the other hand, if the customer makes a direct purchase following receipt of a catalogue, for instance, then we give the credit to the catalogue.
Having got our events and orders lined up, we can start drilling down to different customer segments. For instance, we can look at what is driving repeat customers compared to new customers, or customers from other countries.
Channel effectiveness for returning customers
Using this methodology, we can drill even deeper, and look at particular customer groups such as high value and low value, or those buying particular types of product.
The solution is called UniFida
We’ve been able to automate all of this via a technology we call ADEE which stands for algorithmic direct event evaluation. ADEE gives you the tools to decide how to allocate a marketing budget across multiple online and offline channels, how to focus on recruiting specific groups of customers, and how to encourage existing customers to buy more.
If this sounds a bit geeky, you’ll be relieved to know that all the functionality and AI contained in ADEE is available at the click of a few buttons in the UniFida Customer Data Platform. It has been designed to be used by marketers to help make your decisions simple.
Visit the UniFida Marketing Attribution page
If you would like a demo please email us with a time that works for you.