TFM, Professor Richard Webber and Simon Hay, Segmentation

Sprechen sie segmentation?

Nelson Mandela once said “ if you talk to a man in a language he understands, that goes to his head. If you talk to him in his language that goes to his heart.”

What I love about this quote is its relatability. Through its psychological nature it can be applied to many different scenarios – one of which being marketing.

The important thing about successful marketing is the creation of a positive brand experiences.

I recently gave a talk at TFM alongside Professor Richard Webber, originator of segmentation classifications ACORN and MOSAIC, and spoke about my time at dunnhumby. It is my belief that one of the reasons that segmentation was so successful for Tesco was that we created a common language for the business.

Everyone in a managerial position from the board to the buyers could talk about the customer in the same and unique way.  This meant that the customer was truly positioned at the heart of the business and became entrenched in the DNA. Any decision that was subsequently made was framed around the customer.

From price sensitivity to channel usage, advocacy to coupon redemption Tesco was able to understand them all in relation to how they affected different customer segments. Moreover, because the entire business spoke segmentation we were able to overlay our different segmentation systems to create hyper-personalised customer communications to find cells of just one person creating an extremely effective path to personalisation.

Creating a common language is incredibly important. However, being able to do this is relatively difficult.

Firstly it is important that the segmentation isn’t too complex. The reason that everyone was able to talk segmentation at Tesco was that everyone could understand it. Not everyone has a background in statistics and advanced mathematics and therefore the segmentation system needs to be ‘marketed’ internally.

It can’t just be the domain of the data science team who are able to talk antecedents, permutations, regression and secular trends.  It needs to be humanised – putting the people back into the picture – so that the insight comes alive and is relatable to the business problem at hand.

Moreover, to become a language segmentation needs continuity.

It can’t be changed every month or at the whim of the data science team. It takes time for people to become familiar with a concept and even longer for it to become established. This means that often there needs to be a trade-off between the usability of a system and its level of predictability.

The temptation is to load every piece of data possible into a model in order to predict pretty much anything, but this will make the system too complex and therefore too big for human comprehension and consequently of limited use to the business.

The most useful approach is to create a few systems relating to variables such as clicks, customer value or geography which can be used to provide insight on specific questions such as who are my most valuable clients or where do my customer live, or bring them together to create a three dimensional view of the customer making it possible to determine the most valuable segment of customers by channel and location.

These models are incredibly useful for corporate growth strategies as data science can produce a richer picture of the overall market and specific customer segments.