May 27, 2014 | Written by: Craig Rich
Categorized: IT Infrastructure and Analytics
Share this post:
Why predictive analytics?
In an IDC report sponsored by IBM, it was reported that the median ROI of a predictive analytics project is 250%. Yet, many SAP customers are primarily focused on tackling traditional descriptive analytics projects with business cases based on availability and correctness of information in addition to productivity. While there is certainly business value in these endeavors, IDC estimates the median ROI of these types of projects to be closer to 89%. Descriptive analytics will continue to serve as a solid foundation of any company’s analytics ecosystem, but predictive analytics offers an opportunity to drive additional business value out of your data investments.
Why the difference is ROI?
Predictive analytics business cases are typically based around better decision making and the associated business results. For example – let’s say I want to do churn analysis and predict which customers that are most likely to leave. Reducing monthly churn by fractions of a percentage on a monthly basis can create significant ROI and long term business value. The same can be said for using predictive analytics to drive other decisions such as pricing decisions, inventory decisions, resource allocation decisions, hiring decisions, investment decisions, etc.
SAP HANA Enables Predictive Analytics
SAP is positioning HANA as its advanced analytics platform even though many customers continue to think of it solely as a high performance in-memory database. SAP has quietly embedded 30+ algorithms directly in HANA called the “Predictive Analysis Library” (PAL) which allow developers to apply the most widely used predictive models in business to data directly in HANA. I’ve talked to many customers that own HANA and are surprised to learn they can easily turn on PAL functionality and run predictive models directly in HANA. And if you happen to need a specialized algorithm that’s not part of the 30+ in PAL, HANA also offers R integration with access to thousands of open source predictive algorithms.
How to Get Started With Predictive Analytics Using HANA
- Pick a spot and ask the right business questions that are either difficult to answer today or have non-optimal answers. In the example above, your sales team may have asked for a 360 degree customer dashboard or a set of customer reports in the past. In the predictive analytics world, help them ask forward looking questions such as “Which customer is most likely to leave?” or “What products should I be cross-selling to my currently customers”? This will likely be the most challenging part of getting started because business SMEs often have a difficult time articulating these questions.
- Investigate & prove the value in HANA. If you already own HANA, you already own predictive analytics tools and this makes the business case easier to sell. Start with a 8-12 week proof-of-value justifying that the predictive model “works” and creates business value. Next you’ll need to add in data and resources. This will required some help from someone with a statistical background (IBM can help). You will need to load the data required in HANA and evaluate different models using PAL. But don’t worry about automating data loads, etc. as we’ll do this later once. Plus, we usually learn during the modeling process that we only need some of the data in the final model.
- Operationalize and Productionalize. Once we know the value, we can then decide if it’s worth productionalizing. This means automating the data loads, embedding the model into existing processes, and production deployment.
- Repeat. Congratulations, the first time is always the hardest! Now it’s time to apply tackle those other high value opportunities one by one with predictive analytics. And remember, if data required for future predictive models is already in HANA, it can be easily reused in these new models.