How Marks & Spencers leverages data science to reduce food wastage

How’s the appetite for Marks & Spencer’s food demand forecasts? John Bensalhia finds out from its data science chief, Callum Staff

Hungry for more information about Marks & Spencer’s food data science initiative? Well, you’ve come to the right place.

Heading up the company’s data science function is Callum Staff, who joined in April 2018 to set the team up. “We’re still relatively new, but we’ve grown pretty rapidly, which is great, because to me it’s a really good sign the business sees the benefits.”

Prior to M&S, Callum studied engineering at university, and then was a Civil Servant for four years, starting off on the Fast Stream Graduate Programme at the Food Standards Agency. Callum then moved to the Department of Education, where he set up a data science team there.

“For me, there are three things you need to make sure are in place before you get a data team in, and three things you need to get right to make sure the team is valuable once it’s in place,” says Callum.

“For a good environment to set a team up in, it’s data (having the datasets available for analysis to be done on), projects (making sure the questions that the organisation wants answered are best done through a data science approach), and culture (having the support mechanisms and exploratory approach to problem solving that means data science will flourish).”

“For making sure the team is useful it’s maths (making sure you’re using the most appropriate, robust techniques), IT (having the tools to do the maths and get the data), and culture. Culture is the common theme here, yet getting that culture is the hardest of all three.”

Callum’s latest role at M&S connects with two key principles.

“All of our projects in some way link back to reduce waste and improving availability of product for customers, be that improved demand forecasting or price optimisation reduced products.”

In terms of making food demand forecasts as accurate as possible, the benefits speak for themselves.

“Food demand forecasting, and food supply chains more generally, are a constant balance between not wasting too much food that goes out of date (a timely social issue currently) and not having bad availability of food – empty shelves,” explains Callum.

“Forecast accuracy is only one part of the jigsaw that can lead to waste or availability issues, but having those numbers being correct or at least as close as possible is really important, both from an economic and an environmental point of view.”

Across the sector of food demand forecast technology, machine learning is a key driver.

“There are so many opportunities to apply it to assist both forecasting and retail/supply chain operations more generally, so I think you can expect an increasingly large amount of case studies coming out from the sector as the trend continues,” says Callum.

“Interestingly though, and I don’t think it’s solely the retail industry that is guilty of it, I’m hearing the term ‘AI’ used when it’s really nothing even close to AI (to me DeepMind are the closest thing to a public example of AI right now).”

“We have a *slightly* tongue in cheek ban on the term in my team – we’re working on machine learning projects, and in some cases quite advanced machine learning, but we’re not anywhere near AI, and I wouldn’t want to over-promise and hype by saying we, or others, were.”

Looking forward, Callum is confident that machine learning and data science will continue to spread through the sector.

“I also think that the use of tools and techniques such as graph databases and network graph analysis is really going to take off.”

But one thing that Callum is hopeful for is a closer link between academia and industry in the space.

“There’s a wealth of academic research around forecasting and supply chains and some great opportunities for industry to capitalise on it. We started by doing our little bit last year by hosting two Data Science Masters students to do their dissertations”