Did you know that every year millions of people make New Year’s resolutions? Did you know that less than 10% actually keep them? One of the most popular New Year’s resolutions is to lose weight in the new year. I think of my own experience in the snack food industry seeing a slight drop in demand in January and then back to usual demand by February. But what about your demand planning process. Have you ever thought about trimming that up? Is there excess waste in your process that you could afford to lose? Will 2021 be there year where you have learned from 2020 and perhaps introduce a leaner process?


The first step in any program to improve a process is to look where you are now. I am shocked to hear how many companies either do not measure forecast accuracy or they just do not think that improvements just aren’t capable of achieving. You cannot fix what you do not measure. When thinking about forecast accuracy, I used to think about it across many levels. I was used to companies with over 1000 SKU’s. No one wants to forecast all of them. In a hierarchy approach, we were able to forecast at a level that made sense and allowed automatic disaggregation to forecast at lower levels using proportional factors. The time spent forecasting the SKU level not only improved the forecast accuracy at the category level but helped save a lot of time looking at each SKU.

Also, I recommend measuring forecast accuracy across different channels or customers. What you find is that some channels and customers are more reliable (or “forecastable”). Then you can focus your time and efforts to those that need more attention. I remember holding different forecasters to the same standard. In the core markets and long-time key accounts, the forecast accuracy was fairly good. However, the forecaster from the new market was frustrated that they could not achieve the same accuracy as the core markets. Therefore, you would have a better return of investment of your time spending on customers/markets that needed more focus rather than looking at the overall forecast accuracy.

Measuring the forecasting process is also helpful to improve the process. I recall in one of my roles, we used to measure forecast accuracy at the statistical forecast and each of the overwrites. Sales, demand planner, forecast review meeting, and S&OP meetings, would all have overwritten to the forecast. By capturing the accuracy at each step, we found that the S&OP overwrites were less accurate and added no value. By measuring the accuracy, we were able to bring data to show that the time wasted in S&OP to overwrite the forecast was wasted time that could be spent on more impactful discussions. This could not have occurred if we did not have the data and the forecast accuracy measured.


Demand Variable Index (DVI) and Coefficient of Variation (COV) are two ways of looking at an item as forecastable or not. To calculate DVI or COV, you simply sum the absolute deviation and divide by the mean (or average). Typically, any items that have an index of 0.75 or higher is usually deemed unforecastable. Keep in mind, some items with a high DVI or COV may still be forecastable depending on the item. For example, a new item with fewer data points may have some high standard deviation and may need a different model such as a new product forecasting model. Lower indexes can show lower variability and time series statistical models typically fit. This is helpful with regards to trimming up your process with models or approaches to use for forecasting.

Some items have “lumpy” or inconsistent demand. I am seeing this more and more with the unpredictable customer behavior. This may have “forecastable” DVI or COV indexes and the models most associated with these items may include exponential smoothing, weighted averages, or Croston models. While the accuracy measures may be acceptable, you may have to carry more safety stock or supply chain needs to be aware of the swings in demand.

Communication and Collaboration

I have trained Demand Planners to communicate as much as possible. Demand Planners not only need to communicate the forecast but also the assumptions behind them. Demand Planners need to be vocal and communicate. I have trained demand planners to sit in as many meetings as possible to gather information. We have discussed among many other practitioners the value of “soft skills” for the Demand Planner. Demand Planners should have a good understanding of the business in general. For example, demand planners should be hearing what is going on with marketing, sales, new products, supply chain plans as some examples. I used to encourage Demand Planners to attend as many meetings as possible as information gatherers. This will help the forecast be more accurate with this information that may be unknown to the statistical models. The more trust and confidence the organization has on demand planning, the more trusted the forecast will become.

What will 2021 hold for your organization? I know that one thing that we learned last year is that we need to be prepared for anything. Understandably, some things no one can predict like a pandemic. But these are good ways to start 2021 on the right track and tighten up your demand planning process. Losing the pandemic pounds of old bad habits in your demand planning process will help make the company more productive in its time.