Predictive Analytics as a Basis for Forecasting: The Power of Trends
Predictive Analytics Definition: Predictive Analytics is a business forecasting method that analyzes past and present data and utilizes trend analysis to identify patterns and trends which can then be used to predict future activity.
We've already taken time to glimpse into the crystal ball and peer into the future of forecasting. Along the way, we've found that the traditional multi-year projection is giving way to continuous planning. While this has made businesses more resilient and nimble when dealing with adverse events, this seemingly granular approach to forecasting may undersell the power of trends as a valuable forecasting tool.
"The trick isn't finding the data, it's making sense of it."
Awash in Data, Poor in Information
The past will never tell you everything: all it can tell you is what has been. But this is by no means to say that historical data isn't incredibly useful when applied correctly. The key is knowing what data is predictive and what is noise. FP&A professionals, by the very nature of their job description, are often rich in data and poor in actual information. In this age of advanced analytics and aggregators, the trick isn't finding the data, it's making sense of it.
The key to clarity is to be able to look at your data and find the metrics or data points that offer the most insight. We've written before about the difficulty in selecting the right drivers for your business, but even with the drivers selected, the idea of "importance" comes into play. Your business may be littered with metrics, but not every one of them is going to announce its importance.
Ask yourself: When you look at your historical data, what are the pieces of information that make you say "If I know _____, I can tell where I'm going" or "If I had known such-and-such three months ago, I would have taken the following actions." These are the drivers that, taken in the context of the broader environment, can help you build trends and inform the future.
Where Trends are Born
A trend is an aggregation of meaningful data that suggests or predicts a possible outcome or set of outcomes. Take the bedrock question that tire manufacturers must ask: "How do I predict the tire market?" This can be broken down and answered in several different ways.
For the new tire market, the answer may be relatively simple, as it is based on longer-term contracts and manufacturer projections, with relatively less need to examine larger trends.
Forecasting Demand Before the Need Arises
The after-market demand for tires, however, is less obvious, though no less pressing. Without direct ordering information, to forecast demand for replacement tires before the need arises manufacturers must look at historical data and try to observe meaningful trends that drive demand. Key elements include the lifespan of each type of tire, the average number of miles driven in a given year, external trends that impact tire usage, etc. This is the approach taken by our friend Larry Williams when he was at Cooper Tire.
First, his team assessed how each metric correlated with the target, in this case, say, how closely"miles driven" correlated with historical demand. As it happens, "miles driven" was an excellent predictor of the tire market 6 months later. Creating a trend line of "miles driven" enabled the forecasters to predict sales more than 6 months out. With that as the baseline, the team was able to incorporate other metrics, such as gas prices, income growth, and job creation — all of which will have an impact on how many miles drivers are on the road.
The Power of Trends
That's not to say this prediction is all-encompassing or bulletproof. Take a different example from one of our clients: staffing for a university cafeteria. An FP&A professional, armed with 12 months or more of transactional data, can help predict baseline demand in order to optimally staff the facility to cover that demand. This can cover day-of-week, time-of-day and location elements. Then, other factors, such as changes in enrollment levels, popularity of various locations, and even shifts in meal-time or food preferences can have a significant impact on staffing requirements.
Armed with these trends and ratios, the forecaster is better able to analyze and position themselves in the market: Is a trend accelerating or decelerating? How does it respond to market shocks? What are major changes that can cause demand to implode or explode? And, most importantly, How do I position myself and my enterprise to allow for change in pattern to the trend?