Opinion: The power of big data in personalisation

Personalisation is making its way up the agenda of most gaming companies. There was an article on this very issue in the September edition of eGaming Review. Adam Ruffett (head of casino at Unibet) argued the case for the analysis of the transactional data that customers generate to help tailor their experience with the gaming provider.

So how might we do this? Start with your customer satisfaction tracking data. You probably interview a few hundred people every month about their experience of your company as a gaming provider. If you don’t then you should consider setting this up pretty soon – most of your competitors are already doing this.

In your survey you have quite a few questions about specific areas of performance. Is the withdrawal process easy? Do you have the latest games they want? Can they get through easily to talk to someone when they need to? And you also ask them some overall measures such as how positively they view your brand, if they intend to continue to use you, if they would recommend you to other gamers and so on. Typically you focus on levels of performance on the different attributes, and keep a very close eye on the overall measures such as loyalty and satisfaction, which act as KPIs for the business.

So how do we take this data further to help provide a more personalised experience for your whole customer database? The key to this is a statistical technique called clusterwise regression. This approach identifies groups of customers, amongst your customer satisfaction survey respondents, whose satisfaction with their experience, and future loyalty to your brand, is driven by different elements of the gaming experience. Here is a slightly simplified example based on some test data we explored:

Just a quick look at this summary tells you there are two groups of customers, groups one and three (with low current satisfaction) who are a potentially higher risk for defection. Fortunately, there are tangible things that the provider can focus on to try to reduce risk of churn.

For the first group, the customers might benefit from additional communication regarding the bonus activation process, or perhaps receive a simplified bonus offer. For the third group, their satisfaction would be increased, and risk of defection decreased, by helping them to understand the money withdrawal process more clearly. This example shows us how it is relatively easy to apply this type of analysis of your customer satisfaction data to deliver a more personalised experience to your customers: an experience that in turn will help to drive retention and profitability.

The extra step, that necessitates the big data analysis skills Ruffett refers to in his article, is to then develop the algorithms necessary to tag all the customers on your database with code to identify which customer group they belong to. Accuracy of this prediction will depend a lot on what information you have stored cleanly on your database, and on the steps the analyst has taken during the cluster wise regression process.

However, even with imperfect accuracy, this statistical technique, when driven by a team that also has the requisite data science skills to work directly with your customer database, can deliver a significant step towards personalisation of customer experience.