ITBusiness.ca

How ‘prescriptive analytics’ could automate business decisions

Prescriptive Analytics

By Cameron Graham, TechnologyAdvice

In our last post we discussed what the on-going trend of consumerization would mean for the business intelligence (BI) market, and specifically those working inside it. However, that’s only one of the directions that BI software is heading, and far from the only one. As some companies are reducing barriers-to-entry, others are introducing sophisticated features designed primarily for enterprise companies. One of the best examples of this is the emergence of predictive, and now prescriptive analytics.

When BI software began to gain prominence, most programs focused on descriptive analytics. Descriptive analytics – true to their name – describe existing data. These solutions sort and organize data into groups, and then present a summary of the actions and patterns found in them. This is traditional BI, and for most companies, still the primary focus. It’s estimated that up to 80 per cent of business analytics is descriptive. When you think of companies looking at their social interactions in order to find customer patterns, or pouring over their sales numbers for trends, you’re thinking of descriptive analytics. This is the main offering of companies such as Cirro, iDashboards, and Chartio.

Other data analytics companies, however, aim to go beyond mere description. Using statistical modelling, predictive analytics is helping companies attempt to peer into the future. Instead of presenting past trends, predictive tools extrapolate trends into the future, and provide companies with an idea of the most probable scenarios. Some of the largest names in enterprise BI are now focused on this type of analysis, including companies such as Tableau and SAP. In a recent TDWI business survey, over half of respondents said they were actively investigating predictive analytics. Over 30 per cent said they were already using such technology.

Of course, any probability-based analysis is subject to limitations. Namely, the quality of the data has to be high, in order for the predictions to be accurate. Unpredicted outside forces can also ruin even the best models. Take March Madness for example. Even with a complete set of data about every team entering the college basketball tournament, famed analyst Nate Silver’s predictive analytics model only correctly forecasted one of the final four teams. That doesn’t mean there isn’t value in such models (even that level of insight into March Madness is impressive), it just means that predictive analytics can only show the relative likelihood of an outcome.

 If you think of the different types of analytics as stepping stones, then the final level (for now) is prescriptive analytics. Building off descriptive and predictive models, prescriptive analytics is essentially the “next big thing.” Whereas predictive BI solutions forecast future sales or customer sentiment, prescriptive models help companies decide on what steps they should take going-forward. These solutions factor in potential resource constraints, project requirements, and various market scenarios in order to present businesses with a series of recommended actions.

That sounds futuristic, but it’s really just an expansion of current predictive models. Instead of using only past data and trends to draw conclusions from, prescriptive solutions look at multiple outside scenarios to judge the best way forward. For example, predictive analytics might help your company forecast consumer preferences over the next year. Prescriptive analytics would then allow your company to analyze possible constraints in that market, such as labor shortages, or resource bottlenecks. In theory, the longer you use such software, the better it becomes at recommending future actions.

There are currently only a few companies focusing on this “final stage” of business intelligence. These include TIBCO Spotfire, which features built-in prescriptive capabilities, and Ayata, whose sole focus is prescriptive technology.

Predictive and prescriptive technologies are mostly limited to enterprise-grade software at the moment. However, as the BI market becomes increasingly consumer-friendly, these features should become available to small and medium businesses.

How would your company use prescriptive analytics? Would you feel comfortable relying on algorithms to shape the direction of your company? Let us know in the comments!

Exit mobile version