igor - Fotolia
McDonald's was late to the market with a mobile app. Despite being one of the largest fast-food brands in the world, McDonald's didn't release an app until around three years ago, long after the advent of smartphones.
The app, which has since seen millions of downloads, collects user data to provide customers with personalized deals and give McDonald's a better sense of who its customers are.
Using analytics and AI in marketing
In a virtual session during Google Cloud Next 2020, employees from the fast-food giant detailed how the mobile app, along with analytics and AI in marketing, enabled McDonald's to turn away from mass marketing.
"While mass marketing has been and remains a great vehicle in reaching the most customers, what it really lacks is the ability to target customers with content that's relevant to them," said David Galinsky, director of global customer data strategy at McDonald's.
Personalized marketing, meanwhile, enables McDonald's to understand its customers better.
"With the right data, we can identify the right customers," he said.
The mobile application provides the restaurant chain with a lot of individual customer data, including the time a customer typically comes into a restaurant, the specific restaurant they tend to go to, the types of items they purchase, and how much they spend.
Google Cloud Platform "is a platform that generates insight from machine learning and automates the generation of campaign audiences to a visualization, too," said Andre Engberts, senior technology director at Publicis Sapient. McDonald's chose consulting and technology services firms Publicis Sapient and Capgemini a few years ago to help with its digital transformation.
Engberts noted that almost all the technology McDonald's uses for marketing is native to the Google Cloud Platform.
Different AI models
McDonald's uses five different models to gather insights into its customer behavior.
One is for RFM analysis (recency, frequency, monetary value), an SQL-based statistical model that essentially segments customers by their value. The model uses data on how recently and frequently a customer goes into a restaurant, and how much they tend to spend.
That alone creates about 15 to 20 segments that are the basis for customer campaign targeting, Engberts said.
Another SQL-based statistical model, for product relevancy, uses data about which items customers bought and how often they bought them, and then shows product relevancy for each customer for different categories.
Similarly, McDonald's also uses a product propensity model, which maps every product as a probability of purchase to every customer.
"Even though you may not have bought a product, people like you have bought that product, so we know your propensity to buy it," Engberts said.
The company boasts a customer churn model as well, which predicts the probability a customer will stop going to McDonald's. If the model predicts a high-value customer will churn, McDonald's targets the customer with a high-value offer to try to retain them.
A customer lifetime value model helps McDonald's predict the overall value a customer is likely to deliver to McDonald's over their lifetime.
McDonald's personalized marketing efforts have increased its sales and transaction counts, and using Google Cloud Platform for automation has helped reduce the number of resources the company needs internally, Engberts said.
"Being able to more accurately predict what a customer is going to do is the single most effective tool that we've had over the past five years," he said. "Being able to drive business value, anticipating customer needs, is vital."
Google Cloud Next 2020 took place virtually this year due to COVID-19 concerns. Google has released prerecorded weekly sessions each Tuesday since July 14. The last session will be on Sept. 8.