Originally published October 16, 2006
I’m sure that most readers have heard about predictive analytics – a quickly evolving technology that is starting to find practical business applications in businesses small and large. If you follow the buzz, what’s not to like? The idea of being able to predict future performance based upon past results, within a determined range of certainty, is surely attractive to most executives trying to improve, and better forecast, their business performance. Unfortunately, until recently, the reality has not lived up to the marketing hype. The reality is that there has been potential for predictive analytics, but to realize this potential you needed one, or several, members of your team to have an advanced degree in mathematical modeling as a prerequisite. And oftentimes, the guys with the white coats were not the same guys who understood the fundamentals of your business – so a void existed for the regular, practical application of these capabilities in the daily decisions made by the executives of many corporations.
So what is different today, and how are companies currently taking advantage of this hot area? The difference is in two areas: better (and more realistic) understanding of the capabilities of predictive analytics and business user-oriented developments.
Regarding capabilities, I trust that most recognize that predictive analytics will not place a crystal ball squarely on the desk of information hungry executives. At the same time, users realize that there must be something beyond simple reactive reporting – it is tough to drive your car forward by simply looking in the rearview mirror. Many executives use “gut feeling” when looking at their historic data and estimating future performance. Predictive analytics applies some structure to this approach, enabling executives to make better decisions with a higher degree of certainty. Let’s take a look at three practical (and simple) examples of how predictive analytics helps with this approach: slope, seasonality and correlation.
It is common for reports to show product revenue results, comparing them with product costs to achieve a certain degree of profitability. In many businesses, revenue growth is not linearly correlated with costs, so the profit varies over time. The challenge is to determine the optimally profitable point in the life cycle of your product or service, and try to plan as close to that point as possible, or manage around it. This involves analysis of the “slope” of your profitability line over a range of revenue outcomes. This is fairly straightforward analysis leveraging basic math, but important to note for predictive capabilities. Going one step further may involve managing the seasonality of your business. Many businesses do not have linear demand over the year – and many recognize that their business runs in cycles. It is easy to compare last year’s performance to this year’s projections, but would it make sense to consider data from the last 10 years, and then project forward a model that would consider seasonality with a confidence index for next year’s forecast? Seasonality algorithms address this issue, and there are several with slightly different approaches to the problem. Finally, one additional layer can be applied by tying correlations of performance to other variables in your forecast. Although your business may be seasonal, does the average temperature in 10 major cities also have an impact on your revenue and costs? Integrating this correlation within your forecast model, based upon both the history of this variable as well as relevant future projections, can also improve your confidence in forecast projections.
The second difference involves the notable improvement in terms of ease of use of these technologies. Many thought-leading companies are developing ways of integrating predictive capabilities within their solution set, either directly or through partners. This allows end users to leverage the strength of these mathematical models, often without even knowing they are doing so. Several vendors are already taking this route, and we expect this to grow over time. In addition, some software solutions are making users aware they are using the technology, but are developing simplified user interfaces for access. For example, one recent leader in this area has developed common language access to complex analytics – where users would type in a query that would be of interest to them in common language, and pre-configured reports leveraging sophisticated predictive analytics capabilities come into play in response. Although it is not yet artificial intelligence, it is clearly going down the path of helping users better access information that can help better predict business performance.
In summary, tools are now available that allow executives to go beyond reactive reporting and enable them to proactively predict business performance. The tools are more user friendly than in the past, and this continues to be a developing area to watch closely. Companies that do take advantage of these capabilities will likely find themselves with a significant advantage in better management of their business.
Recent articles by John Colbert