Eric Siegel opened Predictive Analytics World with a keynote on Five Ways Predictive Analytics Cuts Enterprise Risk. Predictive analytics is a hot topic these days, with demand in terms of jobs for instance doubling since the beginning of 2009. Yet the audience was really clear that most executives do not get the true power of predictive analytics so clearly there is more to do.
When people talk about risk and risk management they usually mean avoiding large, catastrophic risks like oil rig explosions. And these risks do matter and these kind of macro risky events do happen. But the focus for predictive analytics is actually on micro-risk, the risk of an individual having an accident or needing healthcare for instance. While this is a common approach in insurance, where the risk of a customer needing to claim is assessed, the approach is actually valid in all businesses. There is a risk that a customer will become a loss, that they will leave, that they will be missed in a targeting exercise or they might commit fraud. Taken together these are a macro risk to a company, but it must be exposed one risk at a time. And this is a critical use of predictive analytics. It is not enough, of course, for a company to simply predict or estimate risk. It must also take actions to mitigate and manage this risk.
Predictive Analytics can be defined lots of ways – I like “predictive analytics turn uncertainty about the future into usable probability” – but one way is to regard it as a data-driven way to manage risk at a micro-level. So, 5 ways to manage risk with predictive analytics:
- Insurance use predictive analytics to augment their standard actuarial models
- Sales and marketing can manage retention risk, purchase risk etc. Churn and response modeling are both part of this, but it is important to use net lift modeling that combines the two – incremental risk modeling
- Fraud is an obvious case, estimating fraud for invoices, credit card, telecoms, paid ads and more.
- Reliability modeling (or unreliability prediction) for everything from satellites to enterprise system components
- Credit assignment is perhaps the classic use of predictive analytics to reduce risk
Of course healthcare risk, education risk (drop out) and more could be added to this list. Eric also touched on the risk of predictive analytics and how to cut those risks (John Elder has a great list of top problems with data mining for instance).
As always with predictive analytics, some of the scenarios Eric discussed are automated uses of models and some are a mix of automation (decision management) and human support (decision support). And the importance of these individual risk assessments when rolled up into a macro-risk assessment cannot be overstated.
Eric recapped by emphasizing the importance of predictive analytics in assessing micro risks and in playing a critical role in risk management. I have written before on the power of analytics to manage risk by risk and this one on micro-decisions.
Comments on this entry are closed.
Good post – thank you! Absolutely right that folks tend to look at the risk of disaster – and don’t recognize what a disaster they can have on their hands customer by customer.
I liked this post I read recently – http://www.pbinsight.com/blog/details/five-common-sense-processes-for-growing-existing-customer-relationships/ – that gets into the risks of keeping unprofitable customers, and of losing profitable ones – not through a crisis, but rather, by not applying and paying attention to predictive analytics.