By: Eric Siegel, Founder, Predictive Analytics World

In anticipation of his upcoming conference presentation, Predicting User and Device Upgrade Issues Moving to Windows as a Service, at Predictive Analytics World San Francisco, April 3-7, 2016, we asked Hans Wolters, Principal Data Scientist, Windows and Devices Group at Hans Wolters imageMicrosoft, a few questions about his work in predictive analytics.

Q: In your work with predictive analytics, what behavior or outcome do your models predict?

A: I have worked in a diverse set of domains.  The predictions ranged from the very serious –  an impending flare for patients suffering from Lupus (SLE) to the more mundane – predicting if players are about to churn out of a social game. At Microsoft, we work on variety of problems, churn prediction is one of them, but my talk at PAW will focus on predicting the Win 10 upgrade experience

Q: How does predictive analytics deliver value at your organization – what is one specific way in which it actively drives decisions or operations?

A: Predictive Analytics delivers values in many ways. It helps with planning and allocating engineering resources and feature prioritization.  It also influences supply chains.

Q: Can you describe a quantitative result, such as the predictive lift of your model or the ROI of an analytics initiative?

A: Often it is hard to quantify this, since you cannot run a true A/B experiment once the model is in production, hence an observed lift can, strictly spoken, not be credited to your model with absolute certainty. I did however in one case observe a 20% decrease in churn by accurately flagging the users at risk for churn  and having a willing partner that implemented retention strategies.

Q: What surprising discovery or insight have you unearthed in your data?

A: I am not sure if it is surprising, but when reasoning about user behavior it is most powerful to look at longitudinal data and then encode changes over time in an appropriate way. The other insight I found is that what users say they do and what you actually observe them doing is most often not that correlated.

Q: Sneak preview: Please tell us a take-away that you will provide during your talk at Predictive Analytics World.

A: One  take-away will be that relying on an insider  population for receiving early signals works well for some aspects and can be very misleading for  others.

Q: What are the most common misconceptions people have about predictive analytics?

A: I encounter most often the perception that building predictive models involves predominantly the application of the latest and greatest ML algorithms. In reality the most effort is spent on data cleaning, imputing missing values and then the engineering of great features. Training an algorithm is the least of our worries.


Don't miss Han’s conference presentation, Predicting User and Device Upgrade Issues Moving to Windows as a Service, on Monday, April 4, 2016 at 11:20 to 11:40 am at Predictive Analytics World San Francisco. Click here to register to attend.

By: Eric Siegel, Founder, Predictive Analytics World