AI in Retail - automating supply chain management and fulfillment at Wakefern and Loblaws

Stuart Lauchlan Profile picture for user slauchlan February 4, 2021
Two exemplars of retailers using AI and machine learning to optimize inventory control and ensure that there's stock on the physical and virtual shelves for customers to buy.

(Pixabay/Company Logos)

COVID-19 and the subsequent panic-buying and societal shift to online shopping put enormous pressures on retail industry supply chain management and inventory control. An increasing number of retailers are looking to AI and machine learning tech to automate back-end processes to ensure that they can always meet customer demand.  

Madeline Bennet’s report on how Ocado is upgrading the robotics in its famously tech-enabled Customer Fulfillment Centers is one high-profile example of how AI can play a critical role in the back-end of retail operations, but the pandemic crisis has accelerated how quickly other retailers are moving along their own automation paths. 

A recent study by supply chain management specialists Blue Yonder - Future of Fulfillment Research Report (registration required for download) - found that more than half (51%) of retailers cited running out of stock as the biggest fulfillment challenge. At present, some 14% of retailers have automation across their fulfillment locations today, with 21% expecting to see full automation in 2021. Nearly a quarter (23%) of retail executives expect to have most of their fulfillment locations automated over the same timeframe, with the drivers for this including improving real-time inventory visibility and orchestration (36%) and better assortment management (36%). 


Among the exemplars of this sort of shift is Wakefern Food Corp, the largest retailer-owned cooperative in the US, made up of 51 member companies which independently own and operate 354 retail supermarkets. With brands such as ShopRite, Price Rite Marketplace and The Fresh Grocer in its portfolio, Wakefern reported retail sales of $18.3 billion for its most recent full year, ending October 2020. 

That was a year in which, according to comments made by Wakefern President and COO Joe Sheridan, a new type customer emerged, one who was looking to cook more at home and shop more online. Meeting the needs of such individuals meant adding capacity to the firm’s online capabilities, including opening a new stand-alone, tech-enabled micro-fulfillment center, its second to date, but to be joined by more in the next few years in order to meet increased e-commerce demand. 

One of the benefits of tapping into AI and machine learning for store replenishment has been freeing up store associates as well as ensuring the shelves are full, says CIO Cheryl Williams:

We have a legacy system that we've had for many many years and it is pretty sophisticated for the system that it is, but what it really required was our stores to understand it and utilize it correctly. What we found with AI is it really doesn't require the skill set necessarily from our store associates. It takes away a lot of the work that they had to do in terms of knowing what levers to adjust or tune to handle things like holidays and first of the month. So [AI] obviously increased our in-stock position significantly and helped those who were not as sophisticated with the tool we had, to really accelerate in terms of stock agility. Another thing that it did for us, is it allowed us to utilize the time that we would have spent on those other things, to really focus on inventory. Inventory is the key to success. If our inventory is correct in the store and we know that that inventory is correct, then the AI takes over and just makes sure that we're in stock.

This also helps with forecasting, she adds, which becomes a kind of virtuous circle: 

A good forecast is going to be dependent upon what the inventory levels are. If we can actually increase the accuracy of those inventory levels through AI, that would be just a totally unique concept. When I talk about implementing AI for store replenishment, people are saving time and when they save the time, they can focus on getting their inventory right. If we can help them get their inventory right through machine learning…we can make sure they are in stock.

In a hugely - and still increasingly - competitive retail sector, having goods on the physical and virtual shelves is critical. Customers will look elsewhere if a retailer can’t meet an order. Equally important is pricing - can a buyer get a better deal elsewhere and how do you as seller make sure that you’re ahead of your rivals here? This was a challenge that Wakefern faced, says Williams: 

Prior to AI and machine learning, I don't know that we had the capability to be as granular as we wanted to be, to go down to specific store pricing for specific items, just because there was not the ability to consume the large amounts of data needed to actually come up with an accurate prediction or projection of what the right retail price should be.

We are heavily promotionally-driven and we have a lot of very unique promotions, so looking at which ones are optimal and drive the most customer value, is something that AI helps us to do. We have been dabbling in a number of solutions and one specifically, that was really done through a test, [means] we are able to churn through our data quick enough for it to make recommendations at shelf level. The engine consumes those changes and continues to learn more.


It’s a similar story at Loblaw Companies, Canada’s largest food retailer, which operates 22 banners, including Loblaws itself. The firm is one that embraces bleeding-edge tech, a recent example being its adoption of a fleet of autonomous grocery delivery vehicles that went into operation last month. Like Wakefern, Loblaw has tapped into Blue Yonder’s AI tech, part of an evolving interest in this field over a number of years. Hardeep Kharaud, SVP Merchandising & Promotions, Loblaw Companies Limited, says: 

We started off with small tactical things that really helped unlock value for people. We found the best adoption [came from] really just making people's jobs easier, delivering better value to them, freeing up time to do other things, but also delivering benefit.  

He cites a case in point of AI usage in the firm's fresh goods division, which started small and grew out: 

The surprising thing for us was how fast people were able to adopt and adjust to it. It was really because of the hands-off approach we were able to have and which made the change management easier at the office, but also in the stores where we work with lots of colleagues. What really makes us a success is seeing the results. When people saw results, the change part of it became easier as well. It became a pretty contagious thing. People hear others talking and that actually helped us accelerate. 

The main benefits of AI in retail at Loblaws mirror the experiences of Wakefern - stock availability, getting inventory in the right spot, bolstering customer satisfaction, cost optimization. Kharaud says: 

Starting with the forecast, but being able to predict ahead of it happening where you have to have the right inventory, is so valuable. From a customer standpoint, being available all the time is increasingly more important, but when you back that up all the way [along] the supply chain to the distribution node, being able to help suppliers understand the needs of the retailer much better and clearer and predicting at a higher accuracy, [that] only adds value. Then that prediction going all the way downstream to all the tertiary suppliers, to the manufacturers, continues to add value. So us being able to predict with even more granularity only adds value. It's an enabler to take costs out of supply chains. 

In terms of pricing, the granularity that’s been seen at Wakefern is also a goal at Loblaws as Kharaud explains: 

With that many outlets and item combinations, people will do a good job, but it's hard to do a great job [with] the dynamic environment, the ever-changing environment. You think about commodities and the price fluctuation you have. People will catch up to it, but it's a much slower reaction pace, whereas implementing technology like AI and machine learning, you can do it at a much faster pace and then be able to integrate that into other technologies that go all the way down to the store, like electronic shelf labels. It's also an area where, if you can build good rules, build good guard-rails, the system can do a good job in helping to make those decisions and freeing up people to do tasks and work on things that are much more valuable and much more rewarding in their day.


As for the future, as the industry looks ahead to operating in the emerging Vaccine Economy, both retailers are focused on understanding what their market looks like post-COVID and how many of the changes seen over the past year will be permanent. For Wakefern, Williams confidently predicts that the online shift is here to stay:

As grocers, the online business was growing and it was growing at a steady pace, but I've read some reports that it's accelerated five years in one year. We've gotten there and I do not believe that customers will start to come back to the stores. I do believe that those who have shifted online will continue to stay online and I do believe more and more people will be going online…We’ve been in e-commerce for I don't remember how long - I feel like 20 years - so it's not like this was a surprise to us, but just the pace that it accelerated [was].

For his part, Loblaws' Kharaud picks out behavioral aspects of the COVID crisis, such as the emphasis on cleanliness and the increase in ‘consolidated shopping’ so as to reduce trips to the store, and wonders how many will stay high-profile. Noting how the events of the past 12 months have led to retailers having to deal with huge levels of volatility, he seems to be expecting the unexpected when he concludes: 

There are no playbooks for this type of stuff.

That said, it’s clear that the continued adoption of AI and machine learning across the retail sector to optimize existing fulfilment options - and potentially create new ones - will be something that can be safely predicted. 

A grey colored placeholder image