New analytics research shows better way to manage inventory

A new model for planning inventory could help retailers account for customers who switch products when their first choice is out of stock.

Product shortages and delays during the COVID-19 pandemic have given everyday people greater visibility into the world of supply chain and inventory management than ever before.

This challenge is particularly important for retailers who deal in "substitutable" products – those that people are willing to swap with a similar product if the one they want is out of stock.

To help inventory managers meet that challenge, recent research by Melbourne Business School Professor of Operations Yalçın Akçay, Deputy Director of the Centre for Business Analytics, with Professor Harihara Prasad Natarajan and Yunke Li of Miami Herbert Business School, is showing a new way forward for managing inventories.

Moving from products to categories

A common performance metric in retailing is to measure the percentage of time a product is in-stock during the selling season. This is mathematically equivalent to the likelihood that a customer is able to purchase their first-choice product.

The more often that a product is available to customers, the more likely that a sale will occur. While most retailers would like to have a stockpile of all products at all times to achieve high sales, this is extremely expensive – and in most cases, unviable.

To manage this tension, inventory planners try to predict how much demand there will be for each product in advance, usually by studying historical sales data.

However, focusing on products individually when they are in a substitutable category could be a costly mistake, says Professor Akçay.

"The actual problem is more complex and would require using advanced analytics," he says.

"Demand for a certain shampoo brand is essentially driven by two factors – there is clearly the demand from customers whose first preference is this brand, but there is also demand from customers who prefer other products in the shampoo category that are currently out of stock.

"It is likely that some of these customers might choose to purchase this product. That is why you need to plan inventory availability for the entire product category in a holistic manner."

A better way to manage inventory

To address this issue, Professor Akçay and his co-authors created a new performance metric. In addition to measuring the availability of each individual product, they recognised the need for a category-wide service level, which measures the likelihood that a customer purchases any product from the category, regardless of whether it was their first-choice or not.

They then created a new optimisation model accounting for uncertain customer demand and stockout-based substitutions, to determine how much of each product in a given category should be ordered from the supplier while simultaneously achieving targeted service levels for products as well as the entire category.

This approach is outlined in their paper Category Inventory Planning With Service Level Requirements and Dynamic Substitutions, published in Production and Operations Management.

"You cannot rely on gut feelings or simple insights to design inventory availability in a product category," Professor Akçay says.

"You need an approach that relies on unpacking the underlying behaviour of customers and their substitution patterns, as well as the intricate interdependencies among various problem variables – demand for a product being driven not only by its own availability, but also the availability of other products in the category.

"On top of that, we don't live in a deterministic world – so you have to take into account the uncertainty of demand."

While the research was based on data from an international IT product retailer, Professor Akçay says the model was built to be scalable and applicable to many consumer-facing vendors and large product categories.

"Using our approach leads to significant cost savings, which of course immediately translates to improved profitability," he says.

"The more willing customers are to substitute products when they go out of stock, the more substantial these improvements will be." 

To read the full research paper, visit Category Inventory Planning With Service Level Requirements and Dynamic Substitutions.

For more analytics information and research, visit our Centre for Business Analytics page.

To find out more about studying at Melbourne Business School, visit our Degree Programs and Short Courses pages, or learn about our range of services For Organisations.