Robert Kugel's Analyst Perspectives

Predictive analytics can be valuable tools for performance management. When the term is applied to planning or forecasting, many people take it to mean the ability to automate plans or forecasts. It’s true that using predictive analytics correctly is likely to enhance their accuracy, but these techniques do not eliminate the need for judgment; in practice, many organizations may realize more value from applying predictive analytics  to assess results than to forecast outcomes. Moreover, as regards performance management the usefulness of predictive analytics extends beyond planning and forecasting. They also can serve to set benchmarks that can be used to assess performance or generate alerts to accelerate necessary action. Although I advise companies to be more aggressive in adopting predictive analytics, I doubt that they will adopt them as fast as they should because of perceptions that the tools are too hard to use and the data too hard to get at.    

That’s too bad, because predictive analytics have many uses in various aspects of business. For instance, as a short-term forecasting tool they can enable a fast-food franchise to project how many hamburgers, fries and soft drinks it will sell in 15-minute increments during the day based on factors such as the day of the week, the time of year, the weather, the volumes sold in the previous weeks, special advertising and promotions and other factors. The owner of the franchise can use the forecast to try to match employee scheduling to demand. For longer-range projections, a builder of Class 8 trucks can use it to forecast quarterly demand based on leading indicators, projected GDP, freight-car loadings, interest rates and other variables found relevant through analysis of historical data. Beyond sales forecasting and budgeting, the company can use these insights to inform its purchases of long-lead items to optimize its parts inventory.   

But as I said, predictive analytics don’t eliminate the importance of judgment in creating plans and forecasts. The analytics rely on historical data and historical relationships, but they shouldn’t be viewed as a black box spewing out unquestioned results. Indeed, these projections depend on a host of assumptions made by the forecaster. Yet applying them properly is a great replacement for naïve extrapolation in that they enable people to do more nuanced forecasts more intelligently. Human beings themselves should use common sense to decide whether the important assumptions apply. If, for example, a demand factor is the average replacement interval, the human mind is the best tool to assess whether this average is likely to lengthen temporarily (say, because buyers are anticipating a new generation of product) or permanently (for instance, because useful lives are lengthening).    

While predictive analytics have value in preparing forecasts, they can be even more powerful as an assessment tool. Especially when businesses are dealing with well-established cyclical patterns (such as a mature, high-volume product like fast food), they can measure results against assumptions. Greater than anticipated burger sales through midmorning may be a reliable indicator that there will be more customers throughout the day and alert the manager of the need to call up additional staff for the dinner hours. In another case, predictive analytics can be used to continuously compare sales volumes from retail scanner data against expected results. A shortfall from predicted sales levels for a heavily promoted item in the first two days of a month-long campaign can serve as a real-time alert. Does parsing the data indicate a specific cause? Further analysis (and the application of human judgment) can determine if there are specific sources of the shortfall. Individual judgment may be necessary because the factor may not be something the internal systems track. For instance, is the shortfall the result of a competitor’s action to neutralize the special promotion? Or is the messaging wrong? Predictive analytics can help companies respond faster to actual outcomes because they can establish reliable performance benchmarks that make it possible to assess results sooner and with greater certainty and thus act sooner rather than later to address an opportunity or issue. This sort of analysis can be applied broadly across the company or to a specific customer. If, for example, sales of letter-size (for the rest of the world, that’s roughly A4) laser paper to Dunder Mifflin is falling below trend, a call to the buyer is certainly in order to figure out why. (Good luck in getting a straight answer there.)    

These examples show that predictive analytics are a powerful tool that companies can use to enhance the accuracy of their plans and forecasts, enable more insightful performance reviews and alert organizations of unexpected outperformance or shortfalls. Yet our benchmark research on analytics in finance shows that only a relative handful of companies are utilizing them. Only 13 percent of organizations overall use them, mainly in specialized functions such as marketing. Just 8 percent of finance departments use them. While my colleague points out that predictive analytics are on the rise, I suspect that the slow adoption of predictive analytics is the result of two main factors. One is that the tools have not been easy for generalists to use since they require training. The other is that it has not been easy to access the data sets. (My colleague David Menninger has written about the IT dimensions retarding adoption here.) I believe these conditions are changing as some vendors try to spur demand by making their software  easier for generalists to use. Still, the issue of data accessibility may continue to impede adoption. Although technical reasons may not prevent a corporation from making needed data more accessible, historically they have shown a general reluctance to make the necessary investments. Moreover, until innovative companies demonstrate clear benefits from using predictive analytics broadly in core business processes, awareness of their potential is likely to remain low, further slowing adoption. Our firm will be research deeper into the adoption and best practices in 2011 but in the meantime, organizations that take the initiative to establish fully functional predictive analytics may put themselves in a superior competitive position.   

Regards,  

Robert Kugel – SVP Research