Customer data analytics, when applied to key profit levers such as cross-sell, retention and cost savings, can help your retail division to increase EBIT (Earnings before interest and taxes) by 10- 25%, which is significant enough to impact share price. The types of data inputs that are relevant to the energy industry will surprise you – including, as I have been recently finding out, dog ownership!
In this article I’m going to give you three examples of how customer data analytics can impact your bottom line (and the relevance of dog ownership):
• Debt provision and/or debt collection
• Field force scheduling, and
• Cross-selling additional products and services.
Data analysis used to improve debt provision and collection
Debt provision is an allowance the auditors demand on the balance sheet to cover aged debt that will have to be written off because at some stage it is deemed uncollectable. Companies debate with their auditors what the value of the debt provision should be but it is typically many 10s or 100s of millions of euros for large utilities. If a company had credible statistics to predict how much of the aged debt it would actually collect, it would have better negotiating power with the auditors to reduce the value of the debt provision. Predictive data analytic models are now being used to provide this evidence and early indications in a water company with several million customers suggests it could result in more than a €10 million one-off benefit to the bottom line.
While improving debt provision simply through better forecasting how much aged debt will actually be paid, can result in a large one-time benefit to the profit and loss account, actually improving collections of bad debt can add a further large and recurring benefit. Early evidence from a proof of concept model suggests data analytics can improve collection of aged debt by half a percentage point every year. In an industry that only makes 4% EBIT this is a 12.5 % EBIT improvement. At least one supplier is offering to build a proof of concept model for under €100,000 that would take just two months to develop and could generate customer microsegments and propensity for each segment to pay aged debt. They could also predict the next best action to take for a customer depending on their recent activity and the microsegment they are in. Using this tailored approach, collections will be improved.
Data analysis used to improve field force scheduling
There has recently been a lot of publicity given to the use of predictive analytics for boiler failures and pre-emptive maintenance. One of the key benefits will be for companies who service boilers annually to cut their field force costs by reducing the visit frequency. While this type of analytics is still in its early stages and can’t be done on boilers of all ages, it turns out that a less headline grabbing modelling technique around boiler breakdowns has been in use for over five years – and is already highly profitable and applies to all boilers of any age. Rather than seeking to predict a specific boiler breakdown, this other model predicts how many will breakdown in a particular geographic area in the next 7-10 days. The output is used to right-size both the call centre and field force ahead of time and leads to lower costs (better planning of staff schedules) and better customer satisfaction due to faster response times.
Data analysis used for cross-sell
Selling additional services can be very valuable. For example, one large European utility is making 0.5% points of its total EBIT from its services business, much of which is cross-sold to its core energy customers.
There are six ways of cross-selling ranging from a toe in the water using affiliate marketing through to doing it all yourself for higher margins. Predictive analytics can make any of these six types of selling more cost-effective. For example, you can improve return on investment of cross-sell campaigns by using propensity-tobuy models to select the right customers for the campaign and to select the optimum campaign size that maximises profitability. In addition to campaign optimisation, the research service covers another four ways in which analytics can improve cross-sell, supported by a huge number of case studies.
How do predictive models work?
A data scientist trains statistical modelling software on historic data, extracts ‘features’ (combinations of data) and finds a best fit formula. The formula is tested on known historic results and the model improved. It is then used on new data to predict future results. The report shows that the model accuracy depends strongly on the data inputs, that unexpected inputs can sometimes play an important role and can be methodically ‘discovered’ (in one cross-sell model for the purchase of PV products, dog ownership turned out to be an important correlation – we can all speculate about the causality, but it’s only the correlation that matters. To best cross-sell PV/energy efficiency products, you want to get hold of data on which of your customers buy dog food) and that there are a huge list of data types you could use and a wide range of sources you can get them from. On the one hand it is quite simple and quick to develop a model that will give good results and fast payback, but on the other hand there is a science and an art of creating best in class models. The market analysis the author has been leading for Delta-ee elaborates on what critical success factors companies should aspire to depending on where they are in their analytic maturity and is the basis for much of the work referred to in this article.
Summary and where to find out more
If data is the new oil – as the expression goes – then analytics are the new oil wells. Without analytics, most of the benefits of all that smart meter and boiler data lie untapped. One thing that strikes me is just how easily some types of analytics can be applied to the entire customer base and just how short the pay-back time can be (often less than 12 months). This is in contrast to some of the connected home business models which I believe can be disruptive but which take much longer and much more visionary thinking to reach the scale necessary to move the dial on overall EBIT.
This article appeared in Metering & Smart Energy International Issue 4 2016.