Putting aside the technical details that only geeks like me enjoy, what is the value to a business owner or department head of spending money on predictive analytics? Why should care?
Simply put: when you spend money on predictive analytics, you are investing in a way to make better business decisions under uncertainty, leading to desired outcomes more often by leveraging untapped value in your data.
- You can only afford to send 500 pieces of direct mail to 5,000 new prospects. How do you make a smart decision on who gets them in order to maximize sales on a fixed marketing budget? With predictive analytics, you can.
- Your employee attrition rate is high. How can you figure out why people are leaving and reduce the cost and consequence of turnover in your business? With predictive analytics, you can.
- Your DSO has been increasing. How do you know which clients to call and ask for payment with the best chances of getting paid ASAP and improving the cash flow of your business? With predictive analytics, you can.
Now there are already lots of ways you probably use data to make decisions…what’s so special about predictive analytics? What’s different is that predictive analytics requires complex algorithms to detect previously unseen patterns in data; patterns that us mere mortals cannot detect with the naked eye, a pivot table and bar chart. These unseen patterns are used to first “predict the past”, and then to help us predict the future.
- Predicting the past entails creating a rule or formula by inspecting actual outcomes in past data to discover a pattern.
- The pattern is not discovered directly by people, but rather by people with special skills applying complex mathematical algorithms to the data, resulting in a predictive rule or formula.
- For a good quality prediction, that pattern should be found in as many actual past outcomes as possible. The more actual past outcomes where the pattern is found, the more accurate your predictions for likely future outcomes will be where that same pattern is discovered.
- Deciding which complex algorithm to run, and how to interpret the results, require new skills and tools that are not likely found in your current firm or organization. You will probably need some outside help at first.
The resultant predictions are statements of probability, not absolutes. Like any statement of probability, their accuracy improves as they are applied to a larger number of decisions. So for the previous three examples, you can provide the following predictions with predictive analytics:
- Based on the pattern discovered in past data, people like “X” are likely to buy the product.
- Based on the pattern discovered in past data, employees like “Y” are likely to leave the company.
- Based on the pattern discovered in past data, clients like “Z” are likely to pay their bill.
And the answers provided by predictive analytics don’t just answer “who” questions. You can also get answers like:
- Based on the pattern discovered in past data, equipment like “X” is likely to fail soon.
- Based on the pattern discovered in past data, projects like “Y” are likely to be profitable.
These predictions about future outcomes are meant to replace and improve upon status quo decision methods such as guessing, gut feel or gurus generalizing from individual experience (this is the story of “Moneyball”, told within the context of baseball.) When you use predictive analytics to make business decisions that lead to desired outcomes more often, you can reduce costs, increase profits, grow revenue or achieve a particular strategic objective important to your business. (In the case of Moneyball, it was winning enough regular season games to make the playoffs, constrained by a low payroll.)
Also, the predictions don’t need to be perfect to be valuable; they just need to be better than status quo decision methods and generate a likely financial benefit that is greater than the cost to do so. Effectively, investing in predictive analytics is akin to building a better mousetrap; in this case, the mousetrap is how you make complex decisions under uncertainty today.
As you might guess, relevant data with good quality and in sufficient quantity is required to make good predictions. But you don’t need big data or big dollars to make good predictions; “small data” may do the job just fine. It really depends on the question you want answered and how effective your status quo decision method is today. Follow the money when deciding where to focus your efforts; small improvements from better decisions made frequently can add up to more dollars in the bank quickly.
Finally, making changes in culture and people is a critical part of getting value from predictive analytics. If you can’t embrace a culture of questions and experiments that generate data for decision making, and if your people don’t trust and won’t use the predictions to make decisions differently, you won’t get the business benefits available from using predictive analytics. All change requires stewardship and management; introducing predictive analytics is no different.
If you’d like explore the business benefits of predictive analytics further, or learn about how other businesses are getting value from predictive analytics, send me an email at firstname.lastname@example.org. Let’s talk!
Gene Connolly is an independent consultant with a passion for data and analytics; reading, writing and sharing what he knows with others who have an interest in how they can harness the power of prediction for themselves. You can find him on LinkedIn , on Twitter or by email at email@example.com.