For any retailer, being out of stock is expensive, and calculating how expensive it is to be out of stock is merely impossible. Obviously, the product that is out of stock is the measurement, but we don’t know if the client wanted to buy much more and left to a competitor where the full shopping took place. Meaning, the retailer didn’t only experience lost sales on the product that was out of stock (OOS), but the whole basket the customer might have purchased. Therefore, the total cost of lost sales will be impossible to know.
Clearly, retailers aim to have the highest product availability as possible. So, if the product availability is 90%, then nine out of ten customers will get the product they’re looking for. Our research shows that 28% finish their shopping without buying the OOS product, about 18% buy from a competitor, and another 17% buy a cheaper substitute product.
Being out of stock is expensive, but it’s also expensive to have too much stock. Although it is common that retailers buy too much inventory for the sole purpose of avoiding being out of stock, that inventory costs money; 1) it’s simply costly to have cash tied up in inventory, 2) old products might have to be sold with a discount, and 3) in some cases the products expire and will be destroyed. Of two evils, being out of stock is more expensive than overstocking, but obviously it’s about reaching the perfect balance between the two.
When analyzing the reasons for being out of stock, there are several factors that matter the most. Obviously, poor demand forecasts count significantly as all procurement becomes imperfect, but there’s another factor that scores higher. Most OOS scenarios are simply due to poor shelf-stocking, i.e. the product is at the back somewhere and the staff has not noticed that it’s finished on the shelf. Fixing those two will certainly increase the product availability rate, but by how much? And what is the current availability rate?
Here’s where another problem starts; most retailers have no clue what their actual product availability rate is, and they don’t know how to measure it correctly. And if the current availability rate is unknown and/or calculated incorrectly, how can retailers know if their supply chain fixes are having any impact? Having more accurate demand forecasts, buying sophisticated automatic store replenishment systems, and hiring more staff for stocking shelves is all worthless if one can’t measure the impact. Most retailers would be shocked to know that their product availability is far less than they actually think.
Machine learning and predictive analytics is proving to be an important part of knowing the actual availability rate and calculating it correctly. By using predictive availability monitoring techniques, retailers can track availability for all products and will not miss any OOS events, and they will actually know for how long the product is out of stock.
For product 1 and 2 it’s a rather straightforward exercise. If product 1 is normally sold evenly throughout the day but one day transactions stop completely around 17:15, the product is deemed out of stock. Likewise, product 2 is also normally sold throughout the day but one day nothing was sold until 13:50, it’s deemed out of stock during that period. This is easy to measure, but the data is rarely as simple as this.
Product 3 shows as a very familiar scenario; it’s sold throughout the day but has gaps in the transaction data. How can retailers know if the product was out of stock or if there was no client buying the product at the time? As mentioned here above, one of the main reasons for OOS events is that the staff failed to notice the empty shelf, causing an OOS event even if the store had inventory at the back. As the transaction pattern for product 3 shows, there is a possibility that the product was not on the shelf during some of the gaps throughout the day, and then the shelves were stocked as soon as the staff noticed.
Knowing if there were no clients buying the product during those gaps or if it was not on the shelf is important in order to measure the correct availability rate. By utilising predictive analytics to understand patterns and behaviour in the transaction data, retailers can accurately predict when products are actually not on the shelves. This practice is fundamental to build an understanding of the current situation and set measures for improvement.
Automatic store replenishment systems can utilize this information for improving the forecast models used for more accurate and timely replenishment. Once retailers can measure the different modifications within the supply chain and what impact it has on OOS rates, they can start improving.
AI is having a major impact on supply chain optimization within retail. As more and more retailers adopt data science and advanced analytics, supply chains improve. And so does the bottom line as detailed here above.