Using AI/ML to Transform Your Retail Demand Planning

I was on a panel last year hosted by AWS Partners Rackspace and retailers Sainsbury’s and Sofology. The final question was, “What are you going to do for the first time in 2021?” As I pondered the question, I thought about what AWS retail customers should do for the first time in 2021. My response to the question: use artificial intelligence and machine learning (AI/ML) in a meaningful way. Most likely, retailers already use AI/ML as part of an existing SaaS solution. However, by meaningful, I meant that retailers should use it in a transformative way to create truly unique customer experiences or optimize critical business processes.

First, Some Context

Let me walk you through some examples of what I mean to give you context. First, let’s look at cost optimization and demand forecasting. The goal is to offer the right quantity of products, in the right place, at the right price, at the right time. You need a sufficient quantity of stock so you don’t sell out of a product in a retail store and lose a sale. But, you don’t want too many items on hand leaving you with an overstock that you have to discount at a potential loss. When you throw in the curveball of 2020, it’s been a challenge to accurately forecast retail demand. As non-essential stores closed, consumers shifted to online shopping in droves. However, retailers that weren’t equipped to fulfill online orders with in-store pick up dealt with huge logistical headaches. Meanwhile, as online orders surged, fulfillment centers didn’t have sufficient quantities of stock to meet order demands. At the same time, buying habits changed. A few quick examples: In fashion, consumers started buying more active wear, or comfortable clothing, to work at home, and in consumer electronics, demand increased for gaming and home computing devices.

Examples of AI/ML to Optimize Demand Planning

Naturally, algorithms can’t solve all of the issues I just mentioned. However, using an AI/ML-based approach to demand forecasting has proven to be very valuable, typically delivering 50% more accuracy over existing tools and practices. And more importantly, an AI/ML-driven approach improves the bottom line, which is more important than ever for most retailers. We’ve seen companies like Tapestry, the luxury fashion house behind brands like Coach and Kate Spade, improve margins through AI/ML-driven stock optimization. In the garment sector of fashion, many clothing items have a max shelf life of 120 days, and we’ve seen customers like River Island in the UK turn to AWS Partner Nextail to optimize in-store stock using AI/ML recommendations to move inventory between stores to reduce stock outs and overstocks. Also in the UK, boohoo, a global brand that’s growing at a phenomenal rate, is using an AI-driven demand forecasting solution from AWS Partner Peak to help its buying teams make better product replenishment decisions.

AI/ML to Avoid Out-of-stock Disappointment

There’s been a tremendous shift in the grocery segment too as online orders have exploded in popularity over the last six months. AWS and Amazon data teams are working with many grocers to use data-only ML techniques to predict in-store stock levels to reduce stock outs and substitutions for home delivery and click-and-collect orders. The algorithms in the software solution track the existing stock levels at each store, and the solution also knows the quantities of items shipping to stores. From this data, the AI/ML-driven algorithms can predict how many consumers will buy a particular item, and the net result is a recommendation for the ideal stock level at any given time of day. Taking it a step further, when a customer is placing an online order for in-store pick up, the AI/ML-driven online ordering system can use sophisticated product recommendation techniques to suppress items with low inventory to minimize disappointment and out-of-stock substitution scenarios.

AI/ML Innovation

AWS Partner Trax is equipping grocery stores with AI/ML-enabled cameras to turn aisles into smart shelves. The cameras monitor products on the shelves to identify low or out-of-stock items and check prices to make sure the correct price or promotional offer is displayed. This real-time insight directs store employees to restock shelves. Meanwhile, the live product data feed improves the demand planning algorithms.

The Meaningful, Transformative Impact of AI/ML

You’re missing the boat if you only view these examples in terms of optimized processes, cost savings, or improved margins. Beyond every one of these initiatives is an improvement in customer satisfaction, and that’s the meaningful impact. If a customer walks into a store, and they are excited to find exactly what they want, in the ideal color and size, you’ve got a happy customer. In the same vein, if an online grocery order is fulfilled exactly as ordered with no substitutions, you’ve got a delighted customer who will return again and again to buy groceries.

I Love These Concepts, but Now What?

AI/ML is an iterative technology that has a cumulative effect. And that means, the sooner you deploy a solution, the sooner you can fine-tune the algorithms to make beneficial improvements to your business. If we circle back to the panel question about 2021 at the beginning of the blog … why AI/ML now in 2021? It’s because tools, technologies, and the AWS Partner ecosystem have matured with proven use cases offering a solid ROI.

If you’re wondering how to get started, it’s easier than you may think. There is a new breed of AI/ML services from AWS that don’t require an army of data scientists, such as Amazon Forecast and Amazon Personalize. Both use proven algorithms from Amazon.com and wrap them in managed services for maximum results. Contact your AWS account team to get started.

Be sure to check out my next blog where I’ll discuss AI/ML solutions to optimize retail supply chains.