Avoid getting out-of-stock: manage inventory using using machine learning

Panic buying during the pandemic hit hard to the stores and millions of the customers across the globe. Thousands of buyers were flocking to the supermarkets and local stores to purchase toilet paper last year, but returned empty handed. The reason: Stockouts!

Being out of stock means that there is huge demand for this product with limited supply. As this product brings good sales in a shorter span of time period but stock-outs cause walkouts. It is the biggest ‘nightmare’ of every retailer.

In this article, we will:

• Develop an understanding of ‘Stockouts’ – an out of stock situation and why is it not good to let it happen in the store
• Dig deep and understand how existing retail and CPG companies are using technology to prevent out-of-stock inventory situation
• Understand the importance of data annotation to prevent out-of-stock inventory situation

A 2018 retail study revealed that the average out-of-stock rate in the US is close to 8 percent (up to 15 percent for advertised sale items).

This Harvard Business Review study highlighted how consumers react to out-of-stock situations:

• Based on product category, 7% to 25% of consumers who face stock-out will continue shopping without buying a substitute product for their desired item.
• 21% to 43% will leave the store and flock to nearby stores to search for the item they want to purchase
• Retailers can lose half of the intended purchases when customers face stock-outs
• Abandoned purchases translate into a sales loss of 4% for a retailer and for retail giants this can be $40 million worth of loss in sales in a year.

To avoid the out of stock situation, retailers and consumer packaged goods companies also try to do overstocking which leads to inventory hoarding, adds up the overhead costs and shrinks margins at the same time.

Out of Stocks impact five key areas:

1. Loss of Sale,
2. Customer Loyalty,
3. Brand Image,
4. Order Fulfillment at Stores,
5. Margin Erosion

Reasons for Out-of-Stock Inventory Situation in Retail and CPG

There are several factors that contribute to the impact caused by out-of-stock like below

1. Unreliable Data: Everything here revolves around inventory data. Shelf level inventory data is often taken during the scheduled timings in most of the stores. Can that data be reliable and accurate? It is apparently not. Because there are items depending on the season of the year, it could be slow moving but high cost item etc. An item may have very low days of inventory on hand (DOH) ratio due to peak sales or customer traffic time during the day.
2. Store Organization: Sales floor and stockroom both affects the customer service and experience. It can test the efficiency of your inventory management system. If the service area inside the store is dysfunctional and unorganized, there is a high risk of items being misplaced. Hence the timely replacement of the item on the shelf can be a challenge. At the same time, the layout of the stockroom affects the inventory turnover.
3. Administrative Failure: Simple administrative and paperwork errors actually account for as much as 21.3% of annual shrinkage — sometimes called “paper shrink.” There are common admin mistakes like incorrect labeling, accounting errors, incorrect markdowns, etc.

Here are some facts about the out-of-stock inventory situation:

• According to a study, empty shelves in the US occur mostly on Friday and Saturday,
• Items on sale have 75% higher chances of running out-of-stock,
• Supermarkets seem to be less likely to face out-of-stock than big-box hubs

Real-World Case Study for Prevention of Out-Of-Stock

The good news is with the advent of Machine Learning, Deep Learning, and Computer Vision technologies, this situation can be avoided. Most retailers are already reaping huge benefits from it. Check them out!

Lowe’s introduced its LoweBot

In 11 Lowe stores in San Fransisco, LoweBot plays a crucial role. These robots create an amazing in-store experience for customers.

If you are looking for anything in particular you can ask LoweBot and it will guide you to the product. Similarly, the robot scans the shelves in real-time and notifies the staff if there is any product that needs to be replenished. Using computer vision they also look for patterns in product or price discrepancies.

Walmart’s Drone Assisted Inventory Management

“The internet of things, drones, delivery robots, 3D-printing and self-driving cars will allow retailers to further automate and optimize supply chains too.” - Doug McMillon, Walmart CEO

Manually checking the inventory can take months together for employees at Walmart. Sophisticated drone technology can complete this task in 24 hours. Drones fly through the warehouse and identify misplaced items and count of units of all the items.

Data Annotation to Prevent Out-of-Stock Inventory Situation

In these modern times, almost all of the data is being captured with the help of monitoring devices and applications. These sophisticated devices enable the capturing of video footage or images of store shelves and warehouses, which is further processed to identify the relevant details.

This processing cannot be accurate if the products are not labeled and identified correctly. That’s why Data Annotation forms a big part of companies looking to achieve automation in the area of inventory management.

Bounding Box annotation is the most popular annotation type used to identify out of stock inventory. Using machine learning, the data captured in-store can be used for multiple purposes including managing out-of-stock inventory.

Every image and video layer captured from the affected area has to tagged accurately. For example, you should be able to determine the detergent brand or baby diaper item, if it is selling out fast. The machine learning model must be able to identify and share an intimation or notification in the real time.

This all boils down to the fact that for any machine learning model to perform its task accurately. the feeding data needs to be reliable, detailed and diligent. That’s where the task of data annotation becomes extremely important.

For this, you require a platform that helps you generate high quality training data for your computer vision team to train and validate your AI applications in a speedy way.

Here is how our machine learning powered data annotation platform at Labellerr helps:

• It enables automated listing of products in inventory,
• It also facilitate automated report generation for quantity analytics,
• To identify the levels of inventory, Labellerr platform uses object detection, image segmentation, dot annotation, and multi-class image classification.

Check out Labellerr’s pre-trained machine learning models for Retail, CPG, and eCommerce that are easily accessible via APIs. Sign up with Labellerr today and see how you can generate high-quality training data to build your AI/ML model faster.

The article first appeared on the Labellerr blog at this link