The brand new Demand Solutions e-book: 42 Principles of Forecasting provides practical and actionable tips for how any company can improve its forecasting process. This article is an extended version of one of those principles: “#7 “Reward Accuracy.”
I used to work in a dairy, a 99-year-old family-run business that manufactured ice cream in every flavor imaginable and every package size that our customers would take. One of my responsibilities as director of marketing was to provide our plant manager with sales forecasts. The forecasts were actually produced by three people who worked for me. None of them were full-time forecasters, but they grudgingly admitted that the word “forecasting” appeared somewhere in their job descriptions.
On the first business day of each month, our forecasting software provided initial projections for each of the next 12 months. I challenged my three forecasters to use their knowledge of our products, customers and markets – as well as their insights into activities like price changes, promotions, new distribution and competitive activity - to shape our system-generated forecasts into more realistic projections. I repeated: “You know things that the system doesn’t know” like an echo.
To encourage at least a degree of attention to the improvement of their forecasts, I held out the prospect of a small reward for improved forecast accuracy. Our incentive program was called “the Jimmy the Greek Award,” and I offered $5.00 each month to the forecaster who most soundly beat the system’s forecasts. (I’m not sure which dates me more damningly – my frame of reference or my penury). If none of them beat the system, they each had to give me a dollar – which I planned to put into a pool and buy lottery tickets as a joint investment for us all at the end of the year.
In all honesty, I really didn’t expect any of them to outperform our software, and I was counting on that lottery pool to fund company-crippling early retirements for the four of us. Much to my surprise however, at least one of my forecasters beat the system forecasts in 23 of the 24 months that we tracked. Much more important, this very small recognition drove a very significant improvement in our overall forecast accuracy. In the first month, our forecast accuracy (based on the reciprocal of Weighted Mean Absolute Percentage Error) was 67%. At the end of two years, we consistently achieved 88% forecast accuracy.
That experience was one of the inspirations for a life-changing decision to start a business to sell and support Demand Solutions - the forecasting software that I first used at the dairy. That was 25 years ago, and in the years that have flown by, I’ve worked with a number of companies that have enhanced their forecasting processes with clever and meaningful incentive programs, built around the performance analysis that they conduct in Demand Solutions.
Several companies developed programs that incorporated the acronym A.C.T. - for Accuracy, Completeness and Timeliness. The A.C.T. approach has worked especially well for companies that called upon their sales force to provide forecasting input. The forecasts had to be reasonably accurate, they had to be submitted in time for the business to put the forecasts to good use, and they had to provide a complete picture of the sales manager’s slice of the business.
Forecasting is a thankless task. You’re always wrong, and your work is never done. Good forecasts aren’t always celebrated, but inaccurate forecasts will always be criticized. There’s a tendency to pin the tail on the forecaster. (I wish I could remember from whom I first heard that phrase many years ago. If it’s you, please let me know!). Poor forecasts deserve to be critiqued, and good forecasts - or consistent improvements in forecast accuracy – equally deserve hosannas and rewards.
In the successful incentive programs that I’ve seen, one absolute key for earning buy-in - from sales managers in particular - was to start with very reasonable expectations for forecast accuracy. For example, in the first several months of the program, if a salesman's forecasts were within 40% of actual sales, they were considered "accurate." The initial tolerances for forecast variance were set to ensure that success was reasonably attainable, and then tightened over time, with forecast error expectations tightened to 35%, then 30%, and eventually 25%. By making it relatively easy to succeed early on, these companies helped ensure enthusiastic participation.
After a trial period, I saw several businesses actually make forecast accuracy a part (albeit a small part) of their bonus plans. Your incentives however, don't have to be lucrative in order to inspire attention and effort. Lottery tickets, lunch with you, tickets to a game or a concert, a gift card, a crystal ball or a wizard's hat on the top forecaster's desk … the recognition is more important than the value of the reward.
Just as incentives work in so many other parts of your business, incentives will also help improve forecast accuracy.