Inventory Optimisation

Published: May 14, 2024
Author: Lorcan Sheehan

What Is Inventory?

Despite being defined as a current asset1, many treat inventory as a liability as attention is usually focused on keeping stock levels to a bare minimum. Of course, this is nothing new. Often, when we speak of supply chain efficiencies, we really mean efficiencies in inventory (more precisely, inventory reductions).

On the other hand, we see a growing emphasis on the costs of not having adequate stock levels. Viewed from this perspective, the impacts of supply chain uncertainties are now decision drivers for many of our clients as they now consider the role of inventory to be a mitigator of risk. In the end, we simply require the inventory control process to provide balance – by maintaining sufficient stock levels to ensure continuous flows while simultaneously keeping inventory levels as low as possible. 

What is Inventory Optimisation? 

Facilitating this balancing of supply and demand is often referred to as inventory optimisation2. Nowadays, optimisation is considered to be the foundation of inventory management – however, setting optimal stock levels can be a formidable undertaking. For one, it involves the interlacing of various analytical methods that enable us make many interconnected decisions to optimise supply chain flows.

Key Principles of inventory Optimisation 

Service and Replenishment

Broadly speaking, the availability of stock drives customer service and usually, fill rate and cycle service level are the preferred metrics used to determine how effectively customer needs are fulfilled. Traditionally, many believed that a high level of stock protected service performance but as consumer demand is now influenced by many factors (environmental conditions, economic, social, etc.), this makes supply chains more dynamic than ever. 

Since over or under-stocking is financially and operationally inefficient, this has motivated many businesses to adopt replenishment methods to balance inventory levels and availability. When it comes to order fulfilment, purchasing decisions are usually made at the SKU-level. Interestingly, now, each SKU generally has its own underlying demand structure, and we now accept that traditional stock policies perform rather poorly when demand is irregular, making new approaches necessary for optimal inventory control.

Classify your Demand 

Understanding your demand patterns is now integral to inventory control; for one, it helps us assess the level of randomness in demand. The principle is this: highly-variable items can be treated in a different way to those that are less variable, and this allows us to match a suitable planning model to a particular item. As a result, we then have a method to optimally tune inventory levels and service performance. 

One popular classification framework3 categorises demand into four groups based on their demand characteristics. When demand is highly regular in quantity and time, variability is low, and the demand structure is Smooth. However, when demand events are regular, but quantities vary, we classify this as Erratic. Then, when regular quantities occur infrequently, demand is said to be Intermittent. Finally, when demand is highly variable in quantity and timing, the structure is said to be Lumpy.

Review your Inventory Levels

All planning models should determine how much and when to order and maintaining a stock review will significantly streamline your inventory management process while the choice of planning method will depend on how often you check your stock levels – these are typically reviewed continuously or periodically. 

In continuous review, we keep a constant track of quantities and as soon as the stock levels reach a prescribed threshold point, an order is placed to replenish the inventory up to a maximum position. On the other hand, for periodic review, inventory is counted at some predetermined moment and an order is generated for an amount that takes the inventory level up to a certain level.

Select your Planning Rules

Inventory control techniques arise mainly from statistical methods that involve probability distributions.  These provide us with the means of calculating the key parameter values for planning execution; re-order points, timing of purchase order placement, economic order quantities, safety stock levels, and so on. 

Many of you will be familiar with the ubiquitous normal rules, and these are found to be best-suited to demand that has Smooth behaviour. However, when non-Smooth demand is encountered, the use of a normal distribution is often not appropriate. 

To model Erratic demand, rules underpinning the so-called gamma distribution are found to be effective while Intermittent demand can be a challenging task in which case, the compound Poisson distribution becomes a model of choice. Lastly, when demand is Lumpy, we find that the negative binomial distribution is a good fit and highly effective in planning.

So, once the demand classes are identified, we can then apply the appropriate planning model to the item in question and in that way, ensure the replenishment cycle is optimised. Over the years, we have applied this approach to a many business environments – thousands of SKUs and millions of units across retail, healthcare, food services, and hardware sectors. Depending on the needs of the client, we generally construct and deploy models that support their inventory management system. Often, we may be asked to provide custom-built solutions in their platform of-choice and this approach has helped our clients drive stock levels down substantially while in the process, improve service performance and reduce working capital and associated expenses including storage, handling, and holding costs.

Manage your Safety Stock Levels

Supply and demand are largely probabilistic, which means there is a definite chance you won’t be able to satisfy some of your demand directly from stock. Holding a level of safety stock is an effective way to absorb volatilities that arise from fluctuating customer demand, forecast inaccuracies and variability in lead times (for raw material and manufacturing).

Provided we define its ideal level, safety stock can then be treated as an asset rather than a liability. Occasionally, some managers prefer to set levels informally and, while easy to execute, this generally results in poor performance over the long-run. Instead, a data-driven approach as described above is recommended to help balance customer service and inventory costs.

Consider Risk Pooling

It is easy to see that if uncertainty could be suppressed, we would require less stock to buffer against fluctuations4. This is the principle behind pooling which states that, under certain conditions5, variability diminishes as demand streams are aggregated across locations, products, or time (think of it this way – it is more likely that high demand from one customer will be offset by low demand from another).

So, when product demands are merged, much of the variability averages-out (giving rise to a narrower distribution of lead time demand) which leads to less risk that in turn, drives a lower level of safety stock and holding costs.

It’s also worthy to note that aggregate forecasts are more accurate than individual forecasts and so, centralising a product in one location provides an immediate advantage. And let us not forget, as stock is consolidated and stored in fewer locations, it becomes easier to manage and the risk of obsolescence reduces. Of course, pooling requires other factors to be taken into account; for example, as stock moves further away from demand, the resulting physical separation between inventory and customer requires strong consideration.

Understand the Role of your Forecasts

There is a deep connection between sales forecasting and inventory management. All forecasts inherently contain uncertainty so, when errors are excessive, stocking levels become affected and ultimately affect product availability or lead to excess safety stocks. 

In many cases, projections are updated weekly or monthly and combined with the lead time and review period to produce a forecast over the protection interval. In part, inventory performance is affected by the forecasting method and accuracy as the timing of when stock should be ordered and how much should be ordered depends on the demand uncertainty.

Do you understand the factors in your business that are susceptible to forecast accuracy? 

To illustrate – in the replenishment of short-life retail products (e.g., food services), forecast accuracy can be crucial to minimising spoilage and waste. On the other hand, longer-life items that move slowly may be more tolerant of inaccuracies. At times there may be little value in constantly tweaking your forecasting models in pursuit of near-perfect accuracy – instead, you might mitigate inherent errors using incremental safety stock for longer-life goods.

When it comes to replenishment, short-term forecasting is at the heart of planning challenges while for non-replenished goods (e.g., apparel), forecasts are often established many months into the future for a selling season that is comparatively short.

Generally, forecasting techniques rely on historic sales data, while others require sales team estimates; in either case, there really is no mystery behind forecasting. Although modern approaches require analytical tools, it’s important to recognise that successful forecasting also requires a knowledge of the market and product domain such as life-cycle position, promotional activity, etc. 

In our experiences, it takes a team of individuals equipped with these skills, working in conjunction with us to develop the most effective forecasting method.

Forecasting techniques range from judgmental (largely intuitive) to quantitative (time-series) methods. 

The one method most associated with inventory management is exponential smoothing which is broadly used for short-term predictions and weighs historic demand so that most of the recent data carry more weight in the moving average. Although we have much to say on this topic, the point is that this technique6 requires the use of smoothing parameters to predict the average demand level and extrapolate for trend and seasonal effects (there are 30 possible variants to choose from). 

In our business engagements, the most compatible algorithms suitable to our customers’ data set is automatically selected and following that, the tuning parameters are optimised to fit the model. We then perform demand simulations before providing our clients with a robust prediction of sales bounded by confidence intervals to help visualise uncertainties and the likely range of future demand.

As expected, products having intermittent demand can pose forecasting difficulties and in such cases, we generally construct specialty models that are modified versions of the so-called Croston strategy.

It sometimes helps to bear in mind the following

  • Forecasts are always wrong
    Forecasts are accurate as long as the actual values fall within the prescribed error range
  • The longer the forecast horizon – the worse the forecast
    It’s much easier to forecast a shorter time horizon with fewer variables than it is otherwise
  • Accuracy improves with the level of aggregation
    See the pooling phenomenon above. For instance – the accuracy of a product forecast at a national level is more accurate than forecasting demand at each individual retail store

What are the Benefits of Inventory Optimisation 

Cost Benefits

In most industries, inventory dominates the annual spend while through discussions, most businesses describe balancing inventory with product availability as a “challenge”.

Properly applied, the aforementioned initiatives should help guide you to unlock inventory optimisation improvements and at the very least, give you the ability to maintain a minimum level of stock and control working capital. 

In reducing your excess inventory and associated carrying costs, you will decrease your operational expenses. In the end, you will also find that this will help reduce your requirement for warehouse space, labour, and other overhead costs (storage, handling, and holding).


Be sure to check out some of our other blogs on the topic of Inventory Optimisation for Service. We also have some industry specific supply chain blogs such as this one that explains the topic of Pharmaceutical Supply Chain Management in 2024!

If you would like to talk to one of our supply chain experts about what PerformanSC can do for your company please get in touch today! 

  1. Generally Accepted Accounting Principles (GAAP): In the sense that it represents potential revenues ↩︎
  2. Strictly, an optimal solution is a decision that provides the best value of an objective function. However, from an inventory management perspective, supply chain managers generally refer to “optimising an inventory control process” when in reality, the goal is to improve the process ↩︎
  3. Syntetos et al. (2005) ↩︎
  4.  In other words – if both lead time and demand are constant, there is no requirement to hold safety stock ↩︎
  5.  Demand streams must be strongly negatively correlated ↩︎
  6.  Holt-Winters’ method ↩︎

All News
Supply Chain Enabled™ - Receive supply chain insights, best practices, and news
Sign up to our communications list