Inventory management methods


Forecast, calculate or simply guess? That’s the choice, that determines whether or not you will have anything to sell. Sometimes intuition helps; but can we always trust it? If not, you have to look for other ways how to properly evaluate demand, in order not to freeze the capital, and to keep the sales growing. So what are the choices? And how to make the right one?


Inventory management – a specific and complex supply chain process. Here, as nowhere else, mathematics and intuition are intertwined, and the proportion of these two elements can often push either towards competitive advantage or financial problems. Since no one in business has learned to measure intuition as a parameter yet, a simpler way to manage stock is frequently chosen, and here mathematics takes up a dominant role.

Currently, there is a range of methods for stock management that help in calculating demand. However, it is not as easy to decide, which one to choose. Moreover, wrong selection or incorrect use may cause severe financial results. Therefore, let‘s take a look at the inventory management methods most frequently met in practice, their application areas, pros and cons.


The human forecast

I chose to begin with the human forecast for its dual character. On the one hand, it is the only method with plenty of space for intuition. But on the other hand, there are areas where this method can be reasonably called the most accurate. It is the only one that allows taking into account the influence of future factors on company‘s results. All other methods are based on historical data analysis. And while there are no algorithms that can predict the future, man is the only one to be able to evaluate events, which did not happen in the past, and will not happen in the future.

In other words, mathematics cannot, while man can predict (or find out) what competitors are planning in the near future, how the changed regulations, legal acts or new taxes will affect marketing, what influence long or short weekends, which change every year, will have, how many potential clients are interested in one or another item, how marketing can be affected by natural disasters and other similar accidents.

The human forecast is mostly used in expensive product markets, where the price of mistake is too high, and where anticipation of a range of hedges to minimize risks is a must. It is also often applied for advance reservations of seasonal goods, when the term until sale is very long. For example, in Northern markets, car battery orders are planned in summer, but their sale depends on whether or not the temperature will drop to -15°C. Who knows? There, you have an example when intuition is applied in practice.

Nevertheless, the human forecast has many disadvantages. First of all, it is the most expensive method, for its main resources are man and information. When company‘s assortment is wide, the ability to predict the effect of future factors on thousands of different items is simply impossible. Or the result will be very inaccurate.

Another problem – not all of market information is easily accessible. In turn, not all of the accessible information is accurate. Finally, internal communication problems are entangled here too, when information doesn‘t reach its users for stock planning in time. Employee’s qualification has a big influence on accuracy of the forecast as well. Inventory planning employees are rarely in the market. And those in the market rarely like to plan inventory.

For all these reasons, the human forecast is mostly used in exceptional cases only, and stock efficiency depends only on employee’s professionalism and experience.


Mathematical forecast

Seeking for more measured solutions can choose mathematical forecasting algorithms. Unlike the human forecasts, mathematical algorithms are always based on historical data analysis; therefore, there is no place for intuition here. Nowadays, a range of traditional forecasting algorithms can be found for inventory planning, and they are often complemented by derivative or complex combinations of algorithms.

Mathematical forecasting algorithms, in one or another form, reflect tendencies of historical data, and accordingly handle situations, where future planned sales basically replicate past tendencies and do not change significantly. In other words, when there are recurrent patterns or evident and stable direction of tendency. However, a specific algorithm is often suitable to reflect only a certain tendency, and its efficiency decreases as sales frequently change. When there are many of such items in assortment, controlling inventory becomes a challenge.

Whenever item’s sales lose pattern, accuracy of mathematical algorithms decreases. This leads to frequent reviewing and changing forecast algorithms or their parameters. Different algorithms of mathematical forecast can differ by parameter sets; therefore, their maintenance becomes complicated. Sometimes respective knowledge in mathematics is necessary for their effective management.

On the other hand, algorithms of mathematical forecast with difficulty interpret exceptions, which often occur in movement of various goods. For example, sales tendencies can be misbalanced by unexpected peak sales, or lack of sales because of item shortage in stock. Similar situations often require employee’s intervention, it is not always simple to identify them; as a result eventually accuracy suffers, and the effectiveness of stock management decreases.


Min-Max, Min or Max

Currently, it is probably the most often method applied in practice. By the way, under its generic name often lies three different types of methods, but they are all connected by threshold value setting principle.

In Min-Max case, two thresholds are set – minimum and maximum. Stock having reached minimum threshold, an order is formed up to maximum threshold. In Max case, only a maximum threshold is set. When item inventory goes lower than Max threshold, an order is formed up to the set threshold. Meanwhile, in Min method case, the stock having reached the set minimum threshold, it is signalled that it is time to form an order, and an employee decides about the quantity.

These methods are popular because they are simple, do not require specific knowledge and are less dependent on the employee’s professionalism. They are used to manage steadily sold item inventory. They can also be useful in situations, when items are sold rarely, but in bigger quantities, and sales intervals are irregular. However, applying Min-Max method for all company’s inventory is often ineffective, and here is why:

First of all, Min-Max method is very static. The established thresholds do not vary depending on changes in sales, so, eventually, it doesn’t protect from potential excess or shortage. If thresholds are set too wide, there might be item excess. If thresholds are too close, potentially, the orders might be formed too often, and that, in turn, increases service costs.

In order to set thresholds appropriately, and in order it would reflect relevant need, they must be periodically evaluated and adjusted. When there is a big quantity of item positions, such work requires considerable human resources. In addition, each adjustment needs a complex analysis and additional calculations; therefore they eventually are less and less reviewed. And that already leads towards financial problems.

One more reason, why Min-Max never fully protects from item shortage, is the programmed irregularity of the method. Depending on sale changes, different item inventory reaches Min threshold at different time. In every such case, a need to order a certain quantity is signalled. However, what to do, if the time for order formation has not come yet? Or what if the necessary quantity is not enough to form an order? Such items are often waiting for a joint order; meanwhile sales are lost every day.


Dynamic buffers

To solve the Min-Max static problem helps method of dynamic buffers, based on E. M. Goldratt’s principles of Theory of Constraints. Generally speaking, it is nothing else than Max threshold (in this method called a buffer) varying in accordance to certain rules. And a set of these rules often determines if the inventory will be managed effectively.

Dynamic buffers partially eliminate constraints of Min-Max methods, because here the Max threshold varies systemically, and its alteration depends on stock alteration. The algorithm itself is complex enough, so its detailed description is not an object of this article, but its simplified operating principle can be described as follows: the stock having decreased too much, if all the rest conditions are met, the buffer is increased by a determined value, in the result of which, a bigger quantity order is formed, and this way the stock is eventually increased. Analogically, if the stock increases unreasonably, the buffer is decreased by a determined value, so the quantity of an item being ordered decreases or the item is not ordered at all, in the result of which, the stock eventually also decreases.

Such automated Max threshold self-regulation principle does reduce the need of human resource, which in other case would be appointed for calculation of Min or Max thresholds; therefore this method is reasonably understood as superior. However, the theoretical possibilities of the method and its practical application very often differ; therefore, there is no point in stating that the dynamic buffer method is enough for an effective inventory management.

The dynamic buffer method can itself be very effective and capable of controlling items with different sales patterns. However, it must be appropriately applied in practice, and in case of this method it is quite difficult to achieve. In order to simplify control, the theoretical possibilities of the method are frequently ignored, so the potential effectiveness of this method is not fully used, and when applied incorrectly it can be harmful.

The main disadvantage of the dynamic buffer method – the complexity of its algorithm and dependence on a range of parameters, each of which can change the stock action unpredictably. To avoid that, a set of parameters depicting the action of algorithm is most frequently not changeable. As a consequence, the company’s entire inventory is by force managed in accordance with the same buffer alteration rules. It would be great, if all items were sold the same as well. Unfortunately, sales of the items differ, and, if managed by the same rules, stock effectiveness might not be increased, but on the contrary – might be lost.

Although called by the dynamic buffer name, the method in practice is controlled statically. The buffer alters by value and periodicity determined in advance. Obviously, this allows to control the tendency of sales alteration only in case that alteration complies with the determined parameter threshold. Otherwise, the buffer will alter insufficiently, too much, too fast or too slow.

For example, when the stock decreases, if other algorithm conditions are met, the buffer is increased by 30%, as defined in the parameters of the algorithm. This fits perfectly for items, sales of which are increasing just as much. However, if item sales increase 15%, why would the buffer be raised to 30%? And what if item sales increase 50%? Or what if the sales increase is only temporary? In cases like this, a responsibility to determine an appropriate buffer is left for the employee. That basically eliminates the versatility of the method and places the decision on the employee’s shoulders. The employee again needs an analysis; consequently, the need of human resources increases and efficiency of the method is lost.

In dynamic buffer method, the action of buffers is influenced by at least 10 different parameters, which in practice are mostly unchangeable, although each of them can significantly affect the buffer’s action, and, respectively, stock effectiveness. To determine an optimal parameter set, more than five billion combinations should be tested with each item, which is impractical and ineffective. For this reason, a compromise is suggested to apply the same parameters for all items. Although this compromise comes at the expense of effectiveness.

Also, the dynamic buffer method is inaccurate in its essence, as it is based on stock change, not sales change. As is known, stock can be influenced by internal transactions, such as item write off or transfer to internal storage to complete temporary operations. In such cases, stock will also decrease, but that doesn’t mean that more of this item should be ordered.

Nevertheless, the dynamic buffer method can be very effective, if there is an opportunity to adapt algorithm parameters for different items and different situations. This method can be irreplaceable in situations, when historical data is not enough to evaluate item’s movement (e.g. when entering new items). It is also applicable in situations, when stock alteration needs to be restrained independently from sales alteration.


Floating buffers

Unlike the dynamic buffer calculation principle, the floating buffer method (moving averages) is not based on stock, but on sales alterations. According to sales data, zone thresholds of safe balance, operating stock and potential demand are calculated separately, the sum of which forms the so called buffer.

This method automatically adjusts to alterations of every item sales, so the buffers alter only as much as it takes to effectively maintain every item’s stock, in accordance with the lead time and order periodicity. In addition, the floating buffer method eliminates most of the parameters used in dynamic buffer calculations; therefore, its control is simple, and it can be easily automated whenever needed.

In case of this method, the buffer can change every day, so the user loses the duty to confirm buffer’s alterations. Not only does it save time, but it also eliminates possible interpretations. And even if the quantity of parameters is significantly smaller, the buffer’s action remains in full control, ant its alterations are adjusted by indicators of alteration intensity and inertness.

Of course, as all other methods, this one also has its own constraints. It needs a sales history in order to operate properly. The longer sales history, the more accurate is the floating buffer method. The more data is evaluated, the more sales alterations of various intensity are traced, and that allows selecting a more accurate intensity of buffer action.


To sum it all up, it is worth mentioning, that there isn’t a single method, which would apply ideally for entire company’s item inventory to be managed just as effective. All methods have their own pros and cons. Some are better for new items, other – for items with steady sales, third – for items with distinctive sales tendency etc. For that reason, stock efficiency is directly dependent on Your ability to select and apply the method, which suits specific case the best.