Forecast ‘error’ affects Inventory in Supply Chains

Roger OakdenLogistics Management, Logistics Planning, Supply Chains & Supply NetworksLeave a Comment

Inventory in warehouse

Not ‘better’ forecasting

Volatility in demand increases as more product lines are introduced that appeal to niche consumers; so ‘better’ forecasting becomes more difficult.

The alternative approach is to analyse how your inventory should be managed. Based on the analysis, build a decision structure for inventory and so better manage the stock keeping units (SKUs).

In the previous blogpost, it was noted that safety inventory is the volume carried in excess of the required cycle inventory (mid-point through the replenishment cycle = half the order quantity). Safety inventory is calculated based on the forecast ‘error’ – a statistical term to identify the variance between planned and actual situations. Under most scenarios, this ensures that customer demands can be met.

Sales forecast as a range

Sales forecasts are often provided for each product group. The planning system should then break down a product group forecast to the SKUs, based on the product mix of prior sales within the group. The forecast for particular SKUs is then adjusted to allow for known sales promotions.

This approach provides one sales number per period (4 weeks or calendar month) for each SKU. However, recognising the volatile nature of sales outside the STEADY category, forecast sales should be identified as a range. This provides an optimistic and pessimistic forecast, with their associated probabilities of each occurring. Multiplying the forecasts by their probabilities and adding the results will provide a more likely indication of future demand.

Forecasts based on a range of outcomes can help the operations of an organisation’s supply chains.. Long lead time items can be purchased using the optimistic forecast level, accepting possible additional inventory. Materials that are easy to obtain on short lead times can be acquired at the pessimistic forecast level. But, orders are placed with the proviso that order amounts can be increased at short notice.

An often used method to measure the ‘forecast error’ of an SKU is the Mean Absolute Percentage Error (MAPE). Logisticians define MAPE as the difference between forecast and actual sales divided by actual sales. This measure can also be used to identify the forecast accuracy for all SKUs in a Category/Class, as discussed in the previous blogpost.

Noting that forecasts are for events in the future and can never be ‘right’, how accurate should they be? For SKUs in the STEADY Category/Class, a MAPE of 10–15 percent is expected. In the VARIABLE and ERRATIC Category/Class, a range of 20–30 percent forecast ‘error’ is likely. The IRREGULAR and LUMPY Category/Class of SKUs can have a forecast ‘error’ up to 60 percent, due to bias in the forecast and unknowns associated with the sales of a SKU.

There may be pressures within your organisation to make the forecast situation look better than it actually is. Two approaches that could be used – but should not, are:

  • limit the forecasts only to the product group level and
  • weight the errors – called the Weighted Mean Absolute Percent Error (WMAPE)

Calculate the safety inventory

Statistically, a 100 percent customer service level can never be obtained. If required you can get close, but it is a high cost, given the amount of inventory that will be held.

To identify the amount of safety inventory for a Category/Class or a SKU, requires a ‘service level factor’ to be calculated. This is based on the standard deviation of the historical data for each SKU.

Inventory safety factor

The ‘safety factor’ is determined from a Positive Z-score chart and the above table is extracted from the chart. This indicates that to achieve a desired customer service level, the standard deviation is multiplied by a factor.

To calculate the standard deviation, each month’s sales variance for the SKU (forecast-actual) is squared and all variances totalled; then divide the total by the number of periods. To finish, calculate the square root of the standard deviation.

The increased costs of inventory and associated areas, to achieve higher levels of customer service illustrate that a Logistician must understand cost implications when making service level decisions.

In addition to safety inventory to cover variations in sales of SKUs, safety inventory is also required to cover the length and variability of lead times for domestic and imported SKUs and materials. This has become more important through the pandemic, as cargo sailings and arrivals have become erratic and more containers are reported lost overboard from ships.

In the time between placing an order with a supplier and receiving the goods, there is a likelihood that actual sales will differ from the forecast. If the interval between placing orders for the item is one month, but the time to receive the order is more than this, the calculated standard deviation must be uplifted by a multiplier, as shown in the table below.

Inventory uplift for standard deviation

In addition, safety inventory is required to allow for variation in the lead time. For each order, identify the variance between the quoted lead time and the actual. To calculate the standard deviation, square the variance and total all the variances for the SKU; then divide the total by the number of orders. The square root of the standard deviation provides the safety inventory for one standard deviation (84 percent service level).

As discussed in the previous blogpost, the Tracking Signal assists with inventory control and action through identifying those SKUs that have a variance outside the target. A control limit is established by Category/Class, calculated by dividing the cumulative variation for the number of periods under review by the standard deviation.

For SKUs in the STEADY Category/Class, the acceptable tracking signal will be up to about 4.0 as the trigger for a review. For the lower categories, forecast error for an SKU will be more varied; therefore, use a tracking signal of more than 7.0.

Not to forget that a critical element in the structured approach for managing inventory is maintaining inventory accuracy. This was emphasised with the introduction of MRP systems many years ago and has not changed, even with the availability of scanners, RFID and bar codes.

The discussion of forecast ‘error’ and its effect on inventory investment illustrates that to reduce inventory requires reductions in variability through the supply chains. This is a major objective for supply chain professionals.

Share This Page

About the Author

Roger Oakden

LinkedIn Twitter Facebook

With my background as a practitioner, consultant and educator, I am uniquely qualified to provide practical learning in supply chains and logistics. I have co-authored a book on these subjects, published by McGraw-Hill. As the program Manager at RMIT University in Melbourne, Australia, I developed and presented the largest supply chain post-graduate program in the Asia Pacific region, with centres in Melbourne, Singapore and Hong Kong. Read More...

Leave a Reply

Your email address will not be published.