M&S actively leverages data science and is eyeing AI to reduce errors, make critical decisions about stock and cut costs via automation.
John Mildinhall, head of data science for retail at Marks & Spencer, told the Retail Technology Show that data science effectively helps with anomaly detection, allowing M&S to pinpoint abnormalities in sales.
This can help identify stores, categories or particuarly products that are underperforming.
Using the existing data and predictive counting, M&S makes accurate forecasts and detects errors at a very granular level, which leads to “much smarter decisions”, according to Mildinhall.
“Data is about relationships. Quite often, you find really counter-intuitive relationships between things in your data. We like to keep an eye on that as well,” Mildinhall points out.
The firm also leverages data science to build product features and gain visibility across all attributes, such as size, description and imagery.
“For example, we are trying to infer body positioning for all all our photos and models to be able to say to a photographer: right, that’s a good pose or that’s not a good pose.”
Subscribe to Charged Retail for free
Mildinhall is particularly excited about the prospects of generative imagery.
M&S spends a lot on photoshop and image manipulation, which can be significantly reduced with large computer vision models, he says.
The same goes for CAD (computer-aided design) data, which can be used to automate the process of designing and documenting a product.
Mildinhall explains: “Say, you a design sofa. That gets turned into a product, then someone writes a description, and then it gets put onto the website, and you have to take a photo. You can automate quite a lot of that along the pathway using CAD data.”
When it comes to artificial intelligence (AI), M&S’ focus is currently on modalities of AI, says Mildinhall. It includes turning large chunks of structured data into sensible reports or finding errors in data sets.
“We’re trying to build an organisation that’s capable of making large learning models and generative models, putting it into our process and actually using it, whilst avoiding the pitfalls. And pitfalls are kind of numerous. Some of them we can’t even anticipate. But we need that kind of level of sophistication to be able to know what happens when it goes wrong,” he said.
M&S, which has over 65,000 employees and operates 1,487 stores globally, has recently extended its partnership with Tata Consultancy Services (TCS) to increase the pace of digital innovation.