Let’s illustrate the complexity with another example similar to one I’ve used before, in which I will once again imagine myself in the marketing department of a mobile phone provider, where I am surrounded by a pretty spectacular BI environment.
This is how my thought process goes:
Already I have a report on the people who ‘churned’ last month (swapped to a different operator), I know who they are, what products they use and when they use them, but I want to know why they churned.
My suspicion is that a new competitor has just started offering a heavily discounted set of tariffs and so I analyse just how much my churned customers could have saved in the last month using this new tariff, and I discover it’s significant but not threatening.
My next task after completing my ‘What’ and ‘Why’ tasks is to try to predict the future. Using advanced analytics and the known profiles of the people that have already churned, I can make a go at predicting who will churn next month, and so out pops my list of likely churners. Now this is a complex task and is never wholly correct, but it’s better than nothing by a long way.
Now let’s make this individually focused so you can see the difficulty and the opportunity. Let’s say that I’ve identified a likelihood that one of my customers, a Jon Page, is likely to churn next month. Now in order to keep Jon it’s very probable that all I have to do is offer him a cheaper tariff, but if I do this, and Jon accepts, what will I have done to Jon’s profitability? The problem here becomes even more complex because I have to do some real thinking. Firstly, do I want to stop Jon churning? If he is in the lowest or non-profitable segment then I should be happy to lose him and so I need take no action. However, if Jon is in the ‘growable’ segment then I must find a way to keep him whilst not reducing profitability over his lifetime. This remains a difficult task involving many cross-relational predictive models and lots of data.
So, let’s imagine that I want to keep Jon without reducing the money he spends on his calls. What can I do? Well, by researching his call behaviour I can identify that he makes many calls on Sunday afternoons at a cheaper rate, so maybe I could offer him these specific calls for free. It won’t make much difference to overall revenue but it might really impress Jon and he might even make up the shortfall by making more or longer calls at another time.
Maybe there’s another way. From the detailed call data I have, I can figure out that in the last three years Jon has holidayed in Malta, so maybe I could offer him a one-off discount voucher for a week’s holiday in Malta for two. At a cost of maybe £200, once and only once, maybe Jon would be so impressed he’d stay a customer for life and he’d certainly tell all his mates how good his provider is!
So, a lot of analysis has to be done, indeed a great deal of thinking and some passion are needed to win.