Combining the “Art and Science” of Demand Planning - Five Top Tips to Improve Forecast Accuracy

 “We need a new system”, that’s what I hear when the topic of Demand Planning and forecast accuracy comes up. Whether it’s a wholesale, catalog, bricks and mortar, or Ecommerce business, the first answer is often the same.  People, processes, and systems, are needed to execute a business plan. I find that in the case of Demand Planning, business process is the area that needs to be addressed first rather than a “new system”. And, if a new system is selected, implementation will require strong business processes, so it’s a perfect place to start for many reasons. Now that you have a well laid out “assortment plan”, it’s time to forecast.

  1. Oddly enough, the first place I’d suggest to begin is to acknowledge that your forecasts will always be inaccurate. Benchmark your industries’ accuracy (or inaccuracy) metrics. How do you compare? Is there realistic room for improvement? Now, make plans (or contingencies) to manage the inaccuracy as well as improve the accuracy. In inventory management this means out of stocks caused by forecasts being too low and “excess inventory” caused by forecasts that are too high as an example.
  2. If you have an idea about what is causing the inaccuracy you’ll be able to address it more effectively. Do you have a way to track inaccuracy? What is causing it? Is it really math/analytics or is it a promotion that the team did not know about? This does not need to be extremely detailed to give you a directional idea where to start. Based on my experience, there is plenty of room for improvement in business processes prior to implementing a new system.
  3. The ‘ole “tops down and bottoms up” really helps. Forecast inaccuracy is larger the further down the merchandise hierarchy that you go (ie sku). At higher levels (ie department), inaccuracy is less. In addition, when you sum “bottoms up” your number usually is too high. “Tops down” along with “bottoms up” is a great way to fine tune your forecast. Often, meeting somewhere is the middle is good. In addition, I always suggest arriving at a number three different ways – perhaps “bottoms up”, “tops down”, and current trend over last year as an example.
  4.  The longest lead time activity should drive a forecasting calendar, whether that’s product placement or marketing development. A calendar should be developed and shared cross-functionally to insure that input is received, forecasting is done, and forecasts are approved in time to meet business deadlines. Often inaccuracy can be improved by simply focusing on timeliness.
  5. Subjective, cross-functional input is crucial.  In my experience this is the biggest area of opportunity in improving forecast accuracy. The team doing the forecast not only needs to know the products, they must know pricing (promotions), space, presentation, marketing activities (advertising, circulation), like items (cannibalization), discontinued items, and trends in the marketplace. Whether you call this an S & OP (wholesale) or the assortment “hand off” meeting (retail), a regular process for obtaining this information is crucial.

Once you have implemented #1-5, you are well on your way to combining the “art and science” of demand planning and will see improvement in your forecast accuracy. You have also put in place items that will be needed if you implement a “new system”. Demand planning systems have nice analytics to help forecast “promotional lifts”, take cannibalization into account, and better forecast seasonality curves. One of the best features of a “new system” is usually the feature to manage by exception which is especially nice if you are dealing with massive amounts of data. Of course, another key element in improving forecast accuracy is to insure the right people are doing the right jobs.

I’d love to hear about your Demand Planning needs and can be contacted at Janice@jlsearsconsulting.com or 206-369-3726.