Playing the Lottery

Back last November when conferences were held in an actual room, I got invited to speak at the ASC conference on "Delivering Better: Turning up the Insight, Value, Speed & Action"

I took the opportunity to showcase some work with a European National Lottery harnessing and integrating both behavioural data and attitudinal data to create a strategic customer segmentation that could be used for both brand/market planning and for direct marketing purposes.

The objective for this National Lottery was to be able to understand their online players from both a behavioural and also an attitudinal perspective. Understand who was playing, what they were playing, how much they were playing and importantly why they were playing. What were their motivations for playing? Did they want to win big or would a smaller win be just fine? What would they do if with the money if they won? Understanding these aspects are key to communicating relevantly on an emotional level as opposed to just a functional level.

The approach was to take transactional and other behavioural data from the CRM and combine that with attitudinal data from a survey of the online players. We selected a significant sample to survey but it was always going to be a sample and so another important critical success factor was to be able to predict which segment a particular player belonged to based purely on their behaviours or demographics i.e. the data that the National Lottery held on everyone. The objective was to be able to predict how people feel about playing based on the way that they play so each player could be assigned to a segment.

Those familiar with segmentation will know that creating the right segments using statistical approaches is as much art as it is science. Segmentations need to be statistically robust but they also need to be commercially robust, they need to make sense and they need to be useful and usable. Without this commercial robustness they won’t get adopted by the organisation and will languish on the shelves.

Blending attitudinal data with behavioural data also brings its challenges. Data collected via surveys for example has a very different shape and feel to digital and transactional data. These behavioural data can bring their challenges in terms of the volume of the data available but by and large it’s scalar and linear whereby 10 is twice the size of 5. Survey based data is more nuanced, if you ask someone to score something on a scale from 1 to 10, a score of 10 isn’t necessarily twice as good as a score of 5 as each person has their own view as to what 10 means. This brings challenges in terms of the analytical techniques used and also the interpretation of the data.

The process yielded 5 segments which differentiated themselves on the key dimensions mentioned discussed earlier, namely what they played, how they played and why they played. In terms of what they played we found that some segments tended to stick to the main draw games such as the Lotto and EuroMillions, others would play more of the sport betting games and scratch cards. Playing behaviour also varied across the segments in terms of the frequency of playing and the transaction amounts. In terms of the attitudes and motivations additional analysis was carried out on the segments to pull out what the main themes appeared to be and we found two key dimensions were around a sense of “me vs we” and whether it was just for a bit of fun or they were hoping that it could be life-changing.

So the segments looked analytically robust and commercially viable. The next stage was to ensure that they were also useful and could be activated by using the behavioural data from the sample only to develop an algorithm to predict what segment anyone in the CRM database was likely to belong to, with initial results showing significant uptick in the test groups.