Melanie Lin, Business analyst ABM Inventory

Proper inventory management is the key to success for any enterprise. Satisfying the preferences of the buyer, minimizing the cost of purchasing, organizing the storage and transportation of goods, and maintaining the optimal stock level are the main challenges here. Though, one should not forget about proper assortment management to satisfy the buyer's preferences, minimize the cost of purchasing, storing, transporting goods, and maintain an optimal stock level so that the buyer can find the goods at any time in the required quantity.

There are different approaches to calculate the stock level at stores (with direct deliveries or the presence of a distribution center - hereinafter DC). Most companies make improvements to their accounting systems to automate orders by choosing the most optimal calculation method for them. Others leave inventory management in manual mode.

Many experts are challenged with choosing the method for managing the company's stock level, especially moving from manual control with a strong human factor to automation. Most often, the company chooses the same method for all types of products. In practice, the success of such a decision depends on whether the chosen method is suitable for all the categories.

## 1. Theory of constraints

One of the most popular and proper methods for managing the inventory level at stores is managing the stock buffer according to the methodology of the theory of constraints. A buffer is an indicative stock level, which is divided into three zones: red (zone of risk of lack of stock), yellow (optimal), green (zone of risk of surplus). Thus, the buffer is changed by 33% up or down if the product dominates in the red/green zone when analyzing the last replenishment cycle.

The method of managing the buffer according to the theory of constraints works best for calculating the needs of stores when the product is not characterized by high frequency and quantity sales variability. A good example is the retailers of the food segment, where the lion's share of the range is occupied by goods with relatively even demand.

An important role in the stable maintenance of the necessary stock level in stores is played by DC, especially if the supplier has a long replenishment cycle. As a rule, the DC delivers goods to the stores at least once a week, and therefore it is easier to calculate the need, and, in case of any miscalculation, to quickly replenish the stock.

Particularly acute is the issue for companies that operate in the non-food segment. Almost the entire range is not a product of everyday demand, and often the replenishment cycle ranges from four weeks and up to six months. That is when DC helps to level the situation in the stores (considering the calculation of the required level of the DC stock is done correctly). In case you deal with high-variability goods, the calculation of the demand for DC orders causes several difficulties since it is impossible to do without a forecast of sales in stores since the DC must replenish the store supply during the entire period before a new batch from the supplier. Not easier is to calculate the stock for own sales of DCs (as in the non-food sector sales are often carried out in the DC).

## 2. Stock level Factors

The calculation of the stock level depends on the demand and the frequency of deliveries. That is why any method comes down to defining how often to order goods and what period to use for analyzing demand. At the same time, the demand needs to be analyzed at each storage point.

The following practices are used:

- In terms of order frequency
- Fixed schedule orders.
- Ordering at the moment of reaching the specified minimum level for the goods.

As for the first method, let's imagine a situation: you have a fixed delivery schedule. This way, it is possible to calculate for which period you need to maintain stock (from delivery to delivery). The more often there is an opportunity to receive the goods, and the less are the conditions for the minimum lots, the more accurate the forecast of the required level of stock will be. Under these conditions, it is easier to adjust to the constant fluctuations in demand: do not pile up the warehouse in large quantities for a long period, but buy in smaller batches and more often in order to be able to quickly restore the goods in case of shortage risk. After all, a large lot can be sold much longer than the expected time, and the return of the goods is not always possible. Moreover, fixed schedules serve as the basis for automating not only the ordering process but also the reporting needed to evaluate the supplier's reliability.

For the second method to work correctly, it is necessary to regularly review the minimum level of goods that may be incorrectly determined or become irrelevant due to changes in demand, and therefore there is a risk of lost sales or, conversely, excess stock.

➥ In terms of demand

In terms of determining the period for analyzing demand, most companies use a fixed period to calculate the average sale, for example, in the last 90 days. Since the company's product range contains products with different sales stability, it is important to structure it by the following criteria:

➥ Sales frequency variability (from one replenishment cycle to another). Variability of sales.

## 3. What is product variability?

Both criteria are determined on the basis of comparative data of several cycles of replenishment. The total variability of the product can be called a combination of two of these types of variability. For its calculation, you can use XYZ analysis, which is based on the calculation of the coefficient of variation of sales. The coefficient of variation is the ratio of the standard deviation of a quantity to its expected (average) value where:

- σ – standard deviation of sales;
- х* – average (arithmetic mean) sale;
- xi – Sales in the i period;
- n – the number of considered sale periods;

The greater the ratio, the less is the uniform demand. When using this method, it is necessary to take into account the availability of goods, because long breakdowns in the supply can significantly affect the result. For a more detailed analysis of the goods falling in group Z, you can make an additional calculation excluding periods with no sales with a balance. This will allow you to select products that can be sold in almost the same volume, but with a large variation in the frequency of sales.

Let us check the relevance of this method of calculating the average sale for a fixed period for goods of different types according to the above-mentioned criteria. Input data: 30 days replenishment cycle, an account of the average sale of storages for 90 days.

➥ Products with low / medium frequency and number of sales are non-variable.

In case of stable inputs, at first glance, it looks like it does not matter for what period to make the calculation. But you need to take into account trends that may change. For example, the first 30 days the product was sold on average of 2 units per day, the following days - 2.17 units, and in the last replenishment cycle, there was an increase in sales of up to 2.5 units per day.

When calculating the average sale for 90 days, this trend will be almost leveled, so it is more relevant to calculate the stock based on demand for the last replenishment cycle or use the calculation of the average weighted sale, assigning coefficients (weight) to each period to take into account the trend in demand.

To calculate the average or average weighted sale, it is important to consider the speed of sales. Therefore, days with a balance at the end of the day or sales per day (if sold to zero) are taken into account. Otherwise, with a long absence of goods, the calculation is likely to be too understated.

➥ Product with high variability of sales frequency.

The calculation based on recent demand for the last replenishment cycle is not appropriate for such products unless it is very long. Example: the product is usually sold at a rate of 4 pieces per cycle, but every 1-3 cycles. Accordingly, it may not be included in the calculation for the last 30 days, but this does not mean that there is no demand for the product at all, and it does not need to be ordered. But calculation for 90 days may underestimate the average sale and lead to lost sales.

## 4. A way out.

In such a case, it is best to add the criterion for dividing the goods into mandatory (primary / high margin) assortment and optional (secondary/low margin). Here you may use a mix of ABCD and VEN analysis.

For example, if a product is significant in the assortment due to margin or the attraction of related products, it makes sense to constantly maintain such a product in a number that is usually sold, despite the fact that it could lie on a shelf for a while. To do this, you can calculate percentiles to exclude explicit sales emissions, but not the usual average of sales, or set up a search for maximum sales during the period. It is this value to be used as a necessary stock level. If the product is not “important”, then it is advisable to calculate the average (arithmetic mean) sales. In both cases, a calculation for a fixed number of replenishment cycles is appropriate in order to capture several periods.

➥ Product with high sales variability in the number.

If the product is sold every replenishment cycle, but with a large variation in the number of sales (this is about unpredictable sales, not promotions/holidays), then using the calculation for 90 days (3 cycles of replenishment) is quite suitable. Please note that if the replenishment cycle is not too long, and coincides with the number of days for calculating the average sale, this can lead to undesirable consequences.

Example: a company uses a method of 30-day calculation, and the replenishment cycle is 30 days. In fact, data is used only for the last replenishment period, and with a large variation in sales from cycle to cycle, this can lead to surplus or lost sales. To avoid this situation, you can determine the calculation period using a fixed number of replenishment cycles.

➥Product with high frequency and quantity sales variability.

This type of product brings the most difficulty in calculating the need. This is a combination of types 2 and 3, therefore, using the method of calculating the average sale for a fixed period, all the difficulties and problems described above may occur.

An important factor for any type of calculation is how often the product is out of stock, as this directly affects the result. For products that do not have days with the balance and/or sale in the calculation, the minimum that can be done is to set the automatic order for 1 unit (items, packaging) or use the previous value. But it is better to set up a notification about positions for which less than n% of days with the balance/sales fell into account.

We found that it is better to use fixed delivery schedules to determine the frequency of orders. What to do with the choice of the period of demand analysis to determine the average sale?

Based on the examples described above, it is wrong to use the same frequency (periods for accounting) for all types of goods. But there is a solution:

Use a fixed number of replenishment cycles to calculate the average sale of goods of uneven demand. In this case, if the goods have very short/long cycles of replenishment, you must specify the minimum and a maximum number of days. For example, if the replenishment cycle is 90 days, and the calculation is made in 3 cycles, then the demand analysis period will be 270 days. In this case, it is harmless to limit the calculated values to six months.

Use a fixed number of replenishment cycles to define the weighted average sales or the last replenishment cycle to calculate the arithmetic average for products according to uniform demand.

So, inventory management is a complex and time-consuming process that should be automated.

What you need is:

- ▷ Analyze and structure the products.
- ▷ Set delivery schedules.
- ▷ Choose the algorithm for inventory management suitable for all product categories.

We covered the methods to deal with products with different frequency, and a number of sales variabilities in case standard practices do not work for you.

*How to deal with goods with low frequency and quantity stability? What if you do not have the opportunity to do last minute orders and receive the deliveries in short periods of time?*