Artificial intelligence and data science are playing an ever-increasing role in decisions by retailers of where to place products in their stores to improve sales.
In years gone by, product positioning was an art practiced by retailers using a combination of data, experience and instinct. Retailers understand their shoppers, the neighbourhood they operate in and the types of households in their catchment area, leading to store layouts that capitalise on the shopping habits of their customers.
But sophisticated data processing techniques and the growth in AI are turning this artform into more of a science, uncovering insights that are not at all obvious at first glance – even to the seasoned retailer.
“We use data to show us patterns using machine learning and deep learning, which the normal human preconceived bias might not find or be able to see,” said Chris Shortt, chief technology officer of Shoprite Holdings, in a recent interview with TechCentral.
“When you bring those two things together, you get the art and the science working very nicely together, and that starts to, hopefully, turn into a better shopping experience from the customer’s perspective.”
Though there are some doubts about the veracity of the story, one of the earliest well-known instances of data processing techniques being used to influence in-store shelving arrangements involves a case study by a now-defunct US retailer called Osco Drug in the 1990s.
Diapers and beer
Shopper receipt information put through a data processing algorithm supposedly revealed an interesting correlation between diaper and beer sales. When analysts examined the data, they hypothesised that new fathers were the link, because they would buy beer on a late-night nappy run, often on a Friday. The assumption was that the beer was added either as a reward or because they desired some well-deserved downtime.
The story goes that the insights were put to the test by putting the nappies as close to beer as possible – next to each other where possible. The result was an increase in the sales of both products.
But retail stores have tens of isles and thousands of products to choose from, which presents a quandary for store managers: how should they organise their store to optimise sales, and how can they experiment with various arrangements without turning the store into a chaotic environment?
Read: SA homeowners turn to AI to fight inflated property valuations
This is where bleeding-edge AI tools shine. According to Shortt, European retailers at the cutting edge of store layout technology are using a combination of IoT, cameras and AI to create digital twins that can be used simulate various store layouts (and their potential outcomes) without having to make any changes to the actual floor plan.
According to Shortt, although the cost of IoT solutions is falling, the cost still prohibits local retailers from using the technology to its full potential. However, South African retailers have a number of tricks up their sleeve to maximise sales by positioning products effectively. The data analysis influencing these layouts begins even before a store is built, he said.
“First of all, we apply the science to say, ‘Where we should even put a store?’ There we’ve got geospatial information, consumer catchment area information and, in some cases, we partner to get data around cellphone coverage in the area. There is also data we can tap into publicly around average income and property sizes. All of this comes together from a profile perspective, telling us what mix of people are in a particular catchment area,” said Shortt.
This broad market data is combined with information about existing competitor outlets in the area to inform the type of store to be built. Shoprite organises shoppers’ behavioural patterns into “missions”, with different types of missions better suited to different types of stores. A cigarette and snacks mission, for example, is better suited to an express outlet, while a month-end family shopping mission is best served by a supermarket like Checkers Hyper.
The core missions a store is meant to serve determines its overall layout, down to the size of its aisles, determined by the number and size of trolleys expected to be on the floor. The general floor plan remains rather static, but items that are on special are usually placed at eye level where shoppers give most of their attention. Items that are usually bought together, like salad dressing and olives, are placed together, too.
Shortt said customer receipt data is an important source of shopper habits, telling retailers which items are bought in which quantities at what time of day. Loyalty cards also allow for the tracking of shopper habits over time, giving retailers a more detailed profile of each customer. But physical store settings have gaps in their data-collection capabilities that the burgeoning online shopping phenomenon is far better at covering.
Curated storefront
“We know what you bought when you check out, but we don’t really know what you did from the moment you entered the store till you got to the till. Which aisles did you browse in? Which product did you pick up, contemplate and then put back down again?” said Shortt.
Apps like Checkers Sixty60 allow retailers to collect vastly more data about individual spending habits, allowing them to predict their customers’ needs accurately enough to make recommendations. The virtual “store aisles” in apps are different for each individual, showing them what they like to see with the option to browse other items in “the store” should they choose to.
Read: The SA start-up using AI to read X-rays – and save lives
“We want customers to choose us for their missions every time, so the more we personalise or hyper-personalise their experience, the more we develop a segment of one where each customer gets to shop in their own curated storefront,” said Shortt. – © 2025 NewsCentral Media
Get breaking news from TechCentral on WhatsApp. Sign up here.