John E. Boylan - Intermittent Demand Forecasting

From a customer perspective, it would be ideal if all of their orders could be met in full, immediately from stock. Consider the example of a single customer order shown in Table 3.1.

There are three levels at which service may be evaluated:

1 Orders: In this example, the order is not completely filled because the order lines for SKUs B and E are not fully satisfied, and the order line for SKU C is not satisfied at all. To calculate the proportion of orders completely filled (the order fill rate [FR]) would require data on all relevant orders. If the order in Table 3.1were included, it would be counted as unfulfilled.

2 Order lines: Two out of the five order lines are completely filled, for SKUs A and D, giving an order‐line fill rate of 40%.

3 Units: Table 3.1shows that 18 out of 25 units are filled from stock, giving a unit fill rate of 72%. Alternatively, this could be calculated according to value. A value of £5100 is filled from stock, out of a total value requested of £6000, giving a value fill rate of 85%.

This example is restricted to a single order, but each of the measures can also be calculated at the aggregate level, over a whole collection of orders. We can find the total number of orders completely filled, the total number of order lines (over all orders) completely filled, and the total number of units filled (over all order lines and all orders). These totals can then be divided by the total orders, order lines, and units demanded, respectively.

Table 3.1 Order comprising five order lines.

To perform well on the order and order‐line fill rate measures requires good service on a wide spectrum of SKUs, including slow‐moving and intermittent items. The unit fill rate is an important measure as it can be applied straightforwardly at both the aggregate and SKU levels. For an individual order line, the value fill rate is the same as the unit fill rate, but these measures may differ when calculated for a whole order, as illustrated in Table 3.1. Full satisfaction of the most expensive order line, for SKU D, has resulted in the value fill rate, of 85%, being somewhat higher than the unit fill rate of 72%.

Although complete order fulfilment is ideal from a client perspective, the percentage of orders completely filled should not be the only service level measure. Retailers usually submit numerous order lines in an order on a wholesaler. Even if the wholesaler improves the availability of stock, this will not necessarily be reflected by the complete order fill rate. For example, if the order lines for SKUs A, B, C, and D were filled completely but only four out of five units were filled for SKU E, then the order would still be counted as unfilled. The order‐line fill rate and the unit fill rate would be raised to 80% and 96%, respectively, thereby reflecting the improved stock availability.

The measures discussed here may be embedded in a broader framework of order fulfilment metrics. Johnson and Davis (1998) described how Hewlett‐Packard augmented inventory holding and fill rate metrics with a measure of customer delivery reliability. This helped to identify the impact of stockouts on delivery delays, including some cases where there was no impact because non‐available stock became available in time for the dispatch of the last truck.

Occasions may arise when an organisation needs to change from one service level measure to another, and new service level targets are required. If service levels have not been monitored according to the new measure, then a relationship between the old and new measures can help to set new targets. Boylan and Johnston (1994) showed how relationships may be obtained between order‐line and order fill rates, and between unit and order‐line fill rates. The relationship between the unit fill rate, , and the order‐line fill rate, , is given by Eq. (3.1).

(3.1)

where is the average number of units unfilled per unfilled order line and is the reciprocal of the average number of units demanded per order line. In Table 3.1, seven units are unfilled over three order lines, giving an average of , to two decimal places. Also in Table 3.1, 25 units are demanded over five order lines, giving an average of and a reciprocal of . In this example, the product of the two ratios in Eq. (3.1)is . Applying Eq. (3.1)to an order‐line fill rate of 0.40 yields a unit fill rate of , agreeing with our original calculation.

Equation (3.1)is useful when the original calculation cannot be performed. This situation can arise when backorder (BO) records are maintained but data is not retained on the quantity filled for each order line. Boylan and Johnston (1994) described such an application at Unipart, a large UK‐based manufacturing, logistics, and consultancy company. In this case, new unit fill targets were determined, based on old order‐line fill targets. This was done separately for each of the company's major logistics clients and for each of the movement classes (fast, medium, and slow moving automotive spare parts) and led to a smooth introduction of the new measure.

In this section, we have stressed that aggregate service and financial targets both need to be attained. Aggregate service level measures may be recorded at order, order line, or unit level and monitored accordingly. It may be beneficial to use more than one measure, so that service can be assessed from a variety of perspectives. If there is a need to convert from one measure to another, then relationships are available for this purpose. In all situations, it is important to gauge the financial implications of hitting service targets, to ensure that the targets are set at an appropriate level.