Analytics That Tell You What to Do

April 3, 2013- 

By Thomas H. Davenport

Most businesspeople are pretty familiar with descriptive analytics—activities that we used to call “reporting” that involve the depiction of data and simple analyses about the past. They are also increasingly familiar with “predictive analytics,” which involve modeling data from the past to make predictions about the future.

Many managers, however, don’t know much about the third category of analytics, which I call “prescriptive.” These are analytics that tell you what to do. They tell you the website version you should adopt, the capital expenditure you should make, the price you should charge, or the optimal route you should drive.

There are a variety of analytical techniques employed in prescriptive analytics. Controlled, randomized experiments are a very effective approach because they are the only real way to ensure a cause-and-effect relationship between the intervention you make and what the outcome is. If your business has a website, you should be familiar with this approach to choosing the best version of it, which is often called “A/B testing” in that domain. Experimental design is also being used in a wide variety of other industries, including banking, retail, hotels, restaurants, etc. If you have multiple locations like branches or stores where you can easily try out different versions of an intervention, this approach to prescriptive analytics is probably a good one for your business.

Optimization is another key prescriptive technique. It tells you the optimal level of one variable in relation to other variables if you want to achieve a particular objective—e.g., making as much money as possible. Pricing optimization is perhaps the most common application of this idea in business, and I’ve always argued that it’s a great way to make money with analytics. Airlines started this a couple of decades ago with yield management, and it’s spread through a variety of consumer and B-to-B industries since then.  Now everybody in the airline industry has this tool, which creates some new challenges for competitive advantage—but that’s a story for another post.

Route optimization in transportation companies is another example of prescriptive analytics. If a company like United Parcel Service Inc. or Schneider National Inc. computes the best route for a driver to take, it can save a lot of money in fuel and time. UPS, for example, has redesigned its route structures based on better mapping data and telematics data from devices in its trucks, and as reported in a recent Wall Street Journal article, “UPS in 2011 reduced fuel consumption by 8.4 million gallons and cut 85 million miles off its routes.”

Prescriptive analytics are great if you actually want to apply analytics to repeated small decisions in your company. However, they are not just an analytics exercise, but rather a change management initiative. Since you’re telling front-line workers—salespeople for pricing optimization, or drivers for route optimization—how to do their jobs, they may resist the use of your prescriptions, or use them in unintended ways. Salespeople might resist optimized prices because they lower commissions. Truck drivers might resist a recommendation from a fuel price optimization algorithm because it directs them to a truck stop they don’t like.

I’m pretty sure we’ll be seeing more prescriptive analytics over the next several years. What that means for managers is that they need to think about the change management aspects of these prescriptive initiatives from the beginning, and enlist the front-line participants as stakeholders in the process. They are all, in a way, “amateur analysts,” and if they understand what’s behind the prescription, they’ll be much more likely to follow it.

Thomas H. Davenport is a Visiting Professor at Harvard Business School, a Distinguished Professor at Babson College, Director of Research at the International Institute for Analytics, and a Senior Advisor to Deloitte Analytics.


On Windows: Navistar Improves Pricing with Big Data


Using Big Data to Optimize Pricing