Thought Leadership July 10th 2019

McDonald's France : business strategy, behavioural analysis & big data

Whether I’m in class or talking to friends, it is often difficult to explain to them what the meta-data in their daily lives can do. It is easy to imagine how big data is used in scientific calculations of meteorological models or for managing transport, but it is harder to see how it can be used in our daily lives.

At first glance, it is simple. McDonald’s is the trailblazer in terms of fast food thanks to its global brand, and its relatively simple strategy i.e. to offer the same quality and the same burgers all over the world. Many of us will have visited a McDonald’s in another country not only because it is easy but also because we know exactly what we will get. However, the competition is growing and customers’ tastes are changing. They want more choice, innovation and change. Competition from Burger King and KFC, but also from new burger chains and artisanal fast food restaurants, are keeping the giant firmly on its toes.

McDonald’s second strategy is “sincerity” as summed up in its slogan “Come as you are”. This burger maker’s second marketing concept gives it an identity based on family values, mates, trendy friends... aren’t we all “friends” on Facebook? And this is where the concept of Big Data comes into its own: behavioural analysis.

But going beyond effective slogans, how can McDonald’s know if you visited one of its restaurants as a family, with friends after drinks or the match, alone with your child at the weekend, etc. Until recently, a customer-centric slogan was fairly meaningless in practice as it was impossible to verify and direct digital (and physical) marketing resources into the brand's Fast Foods.

This is where Romain Girard from the Business Insight division of McDonald’s in France stepped in (2 million visitors/day, €5 billion annual turnover and 1,450 points of sale). He explained that initially the company only analysed its sales receipts for stock replenishment purposes i.e. the number of burgers sold. These receipts were sent to the parent company in France in order to manage stock levels. But McDonald's success is not all about burgers, it is also about understanding, knowing, and clarifying consumer behaviours. Historically, the company looked at sales in a fairly simple way i.e. by point of sale, product, time and type of protein. We have been able to combine all this data. However, it is not enough to know whether a customer visited a restaurant alone, in a group or with family. So, the Insight team analysed the potential of their tools to better determine consumer behaviour in their company. “When we started the project, we held a series of business workshops with the market research, IT and marketing departments to map and audit all our available data,” explains Thibault Labarre from Ekimetrics, a worldwide Data Science consultancy firm. More specifically, data scientists receive terabytes of receipts (two years’ worth).

And this is the heart of the matter: what can we learn from a sales receipt? My extra ticket? On its own your ticket is not worth much to a burger seller, although it can - if you paid with a credit card - tell where you live, where you have come from (if you bought another product minutes or hours before), compile all the McDonald's purchases made on your credit card for several years, etc. But all this is not enough, we still have to delve deeper into the issue by grouping and cross-referencing purchases. Two employees explain that they have hundreds of millions of transactions. Whether it concerns one, two or three people or more. Whether the food was bought in a Drive-In, a restaurant or as a take-away. Whether the customers are known (NDR: via the loyalty card). This represents an incredible wealth of data [...]. These receipts were being underutilised due to a lack of calculation capacity. Therefore, the big data equation is simple as long as you have computers capable of calculating and analysing two terabytes of data and the six billion lines that the last two years’ worth of receipts represent.

The goal: to create a customer typology

“We have receipts and we want to define behaviour segments with them. Put this way, it seems simple. But this approach risks being completely disconnected from the reality and the market research [carried out by McDonald's],” recalls Thibaut Labarre. Romain Girard confirms this and explains, “When you visit a McDonald’s, you can go alone and order yourself a Big Mac. At the weekend, you might visit with the kids and also order a Big Mac. After drinks or a football match, you and friends might visit the restaurant and you also order a Big Mac.” In other words, the information about a Big Mac sale on a receipt does not give any indication about the purchasing context. It is the same product but purchased on different occasions making it hard to analyse the sales.

"For example, one of the questions asked was “how can I know whether it was a family [that visited]?” We had to use a proxy. A “family” is a sales receipt showing a Happy Meal purchase. It is not perfect but it does give a workable result. It also shows the precise forms that the panels can take based on the data on the receipts. With the panels we can know the exact proportion of people who visit a restaurant as a group but pay individually. We can come up with approximations. This means that we can manage proxies (biases) [in this case]. The results are already pretty impressive. We have identified nine typologies (alone, group, family, etc.) allowing a multi-scalar interpretation (from more detailed to more general). But what is truly remarkable, as mentioned earlier, is that today’s calculation capacity allows this type of analysis to be completed in just three months!"

Test & Learn 

OK, that’s all very well I hear you say but... that's why the process McDonald's has put in place includes a continuous learning element in order to refine behavioural analyses. So, as of today it will be far more relevant to read your receipts in terms of targeted marketing i.e. company by company, place by place and, undoubtedly, based on a city’s calendar of events. One thing is certain. If tomorrow the French football team is playing then burger ads will be less generic and more linked to the places where (we now know) groups of people are likely to go to the match or watch it on TV in a neighbourhood with a McDonald's restaurant. And how do we know all this? Because they have already visited a McDonald's restaurant. Therefore, we are very close to direct and targeted marketing... but without your knowledge as it is your sales receipt that has betrayed you. For example, explains the journalist, if we know that after a rugby match groups of friends usually order Big Mac meals, beer and hamburgers then McDonald’s would be wise to discount these meal deals during the Six Nations to attract supporters rather than highlighting Sundays or coffee. The big difference is that this type of promotion will only happen in the right places.

“What is also certain is that the ads you see in the coming months on your TVs and phones will be based on this data,” smiles Romain Girard.



Article published by Numeric Landscape, also available following this link. For more information:

Thought Leadership June 19th 2019

How to combine business approaches, advanced statistics, and technology?

It is sometimes difficult to create a link between highly sophisticated statistical approaches and business reality. The transformation of data into value is an art that needs to combine three different pillars: business understanding, advanced statistics and technology. This triangular approach needs to be adopted by all machine learning projects.