Predictive Analytics

McDonald’s Goes All-In on Machine Learning

By Sriram Parthasarathy
Share on LinkedIn Tweet about this on Twitter Share on Facebook

McDonald’s recently announced that it acquired analytics firm Dynamic Yield, a machine learning and personalization platform for retailers.

The acquisition is the latest data and analytics venture from McDonald’s, which has been investing heavily in digital transformation since 2015. The company paid $300 million for Dynamic Yield—which may sound big, but it’s nothing compared to the fast food chain’s market capital of $143.5 billion.

Now the biggest challenge for McDonald’s is to bring this new technology to their 38,000 stores all over the world. I’m assuming they’ll start with their 14,000 U.S.-based restaurants, but we’ll see.

McDonald’s bet on machine learning and artificial intelligence (AI) is a great sign for the democratization and acceptance of AI in general. Hopefully it will spark other retail chains to consider the possibilities of AI for their businesses, which in turn will create more demand for AI solutions and drive adoption of these tools.

Data Volume and Augmentation

The sheer volume of McDonald’s data is amazing. Every day, 68 million customers visit one of McDonalds’ 38,000 retail locations—and the majority of them do not get out of the car. So, the question becomes: How does one service these drive-in customers with more AI-driven personalization?

Consider the data McDonald’s can use to improve personalization:

  • Historical sales data at each of their franchises
  • External augment data such as weather, traffic, nearby events or activities, and Census data
  • Day of the week/time of day stats
  • Customers’ past purchases
  • Trending items
  • Location information

What can McDonald’s do with all this information? I can see a few excellent use cases:

  1. Promotions: What other products can we recommend to this customer based on their demographics and past purchases?
  2. Real-time recommendations: If a customer purchases this, what other product can we cross-sell? For example, say a customer buys a kid’s meal. Maybe we can recommend a coffee or a snack for their parent.
  3. Customized menu for each patron:There might be a way to identify the customer who is driving in in order to provide a custom personalized menu for that patron. Maybe it’s based on geo-fencing in their app, or on identifying the license plate number using image-based deep learning algorithms like Convolutional Neural Networks (or CNNs).
  4. Dynamic menu changes based on demand: If the line is moving fast, maybe change the menu accordingly. If the checkout line is long, maybe change the menu to only items that are faster to prepare.

This system is going to get smarter and smarter over time. It is learning what recommendations are working and not working. Consequently, that data can be further used to tweak the model and enhance the personalization features.

Packaging the End-to-End Solution Is the Key

The hard part isn’t building the above predictive models. It’s taking all the individual pieces—from raw data to cleanup, to building out the model, predicting, and distributing the predictions to user interfaces, and letting them take action. How the company packages the end-to-end workflow to make use of data and create intelligent predictions that drive actions—that is the value they can create.

Wake-Up Call for Retail Chains

Once McDonald’s starts rolling out this personalized behavior, and assuming it works well, it’s going to make customer ordering so easy. Customers will understand how to use AI-enabled systems and see the value of machine learning—and soon, they’ll start demanding similar systems from other brands. With more and more retail chains beginning to look for such solutions, this move is an early indication of how AI will revolutionize the retail space.


Originally published April 2, 2019; updated on July 31st, 2020

About the Author

Sriram Parthasarathy is the Senior Director of Predictive Analytics at Logi Analytics. Prior to working at Logi, Sriram was a practicing data scientist, implementing and advising companies in healthcare and financial services for their use of Predictive Analytics. Prior to that, Sriram was with MicroStrategy for over a decade, where he led and launched several product modules/offerings to the market.