Part 2: Unlock All that Cash Tied Up in Inventory - Forecast Bias

During these challenging times, many organizations are experiencing cash flow pressures. For that reason, I have published a list of 50 practices that will quickly and sustainably reduce inventory levels.?Quickly?because they do not require investment in expensive technology. These tools focus on using your head, not your wallet.?Sustainably?because they are all about improving the?"critical few"?process gaps, not short term gimmicks, such as refusing supplier deliveries at quarter end.

The list can be found at my website:?https://steveclarkeconsulting.com/papers

Today, I have selected Forecast Bias which unnecessarily causes either excess inventory or inventory shortages, depending upon in which direction the bias exists. This is the worst type of forecast inaccuracy, because it is induced by management practices. Bias is when the forecast is consistently over or under actual demand. The other type of forecast inaccuracy is random variation, in which actual demand fluctuates above or below the forecast each month. Random variation is natural and can not be eliminated, but it can be reduced through process improvements.

Why does forecast bias exist?

One reason is that salespeople tend to be optimistic as a general rule, which can cause them to see future sales through rose-tinted glasses. I find this especially to be the case for new product introductions, where forecasts can be wildly optimistic. In addition, if there is no mechanism to differentiate internal and external forecasts, which have been given to Wall Street, the commercial department could be loath to make necessary reductions to forecasts due to the impact it could have on stock price.

This can also be the case for a start-up organization that is striving to raise capital. One manufacturing company gave incentives to its sales team to obtain orders from its distributors, thereby increasing the backlog, which was a signal to investors that demand was very healthy. However, in reality inventory was building up at the distributors, until the point that their warehouses were overflowing, and they could not accept more deliveries. The channel was well and truly stuffed! Inevitably the distributors cancelled open purchase orders to the manufacturer, which had ramped up capacity in line with the large distributor demand it had recently experienced. Once reality set in, production rates had to be slashed and many employees were laid off.

Another cause of forecast bias is when the sales team does not trust the supply team to consistently deliver, then they may submit aggressive forecasts to increase the likelihood that product will be available when required. In these cases, forecast bias will create excess inventory, and ultimately obsolescence and expired materials. Since scrapped inventory is typically an important metric within the operations environment, they understand the game that is being played, and may choose to disregard the sales forecast, and create their own.

Forecast bias can be in the opposite direction, otherwise known as “sand-bagging”. In this case, forecasts are consistently lower than actual demand. Management often inadvertently encourages this behavior by setting quarterly milestones and bonus plans based upon sales forecasts, which pressures the sales team to submit lower forecasts. In this case, constant inventory shortages will occur, thereby impacting deliveries to customers.

As you see, forecast bias is very disruptive, and the worst part of it, is that it is completely within management’s control. Therefore, the target for forecast bias should be ZERO!

How is forecast bias calculated?

My apologies, but I must introduce a little mathematics here, so hold on tight!

Generally, Running Sum of Forecast Error (RSFE) is used which sums the last 12 months of forecast error at the product family level. As you can see below, the total RSFE over the 12-month period was 17,719 units, compared to the actual demand of 309,887 units, which is approximately 5.7%. Is that too much? There is a mathematical method answer to this question, by calculating the tracking signal.

The tracking signal compares the RSFE to the amount of random variation, which is calculated using Mean Absolute Deviation (MAD). This is a way to normalize the forecast error. In this case, it can be seen that MAD is 2951 units per month. Statistically, forecast bias is significant if it is more than 3 times MAD. The red lines denote the high and low tracking signal, and so you can see that there is a significant bias to the forecast. In this situation, it would then be good practice to drill down further by region, sales territory, SKU etc., whatever makes sense for your organization, to identify the source(s) of the bias, so you can better target how to correct the issue.

I hope you are still there, and the brief mathematics lesson did not scare you away. The broader point here is that forecast bias should be measured, and if it is found to exist, it should be driven out of the process. If removing the bias is not possible, at least in the short term, then it is best to utilize a bias adjustment factor. In this case, the forecast utilized for supply planning would be reduced by 5.7%.

If you have any questions, please contact me at: [email protected] or 415-246-2938