Starbucks: From Coffee Machines to Machine Learning
With more than 20 million regular users on its US app, Starbucks collects more data in any single day than it serves Latte Machiattos. The result is a highly personalized, well-designed customer experience: “With whipped cream, as usual, I assume? Have a great day, Stéphane!”.
Zooming in on the secrets of a lucrative data company that happens to make coffee.
A quarter of Starbucks’ 100 million weekly transactions happen through its mobile app, and that phenomenon has accelerated due to social distancing guidelines during the pandemic. Created in 2011, the Starbucks app was the company’s point of entry into big data and data analytics. Over the last decade, the company’s uses of data have been dramatically diversified to create the most value for its consumers -regulars or not – ultimately strengthening Starbucks as an unbeatable leader in coffee shops.
The Starbucks app collects and uses data in different ways. Its first use was to create a loyalty program, where consumers could collect ‘stars’ for every purchase, and redeem them on selected drinks. The app quickly became a convenient hub where users could access information about menus, store locations, and opening hours. This provided the company with useful information about the most popular store locations, drinks, and times of the day.
But Starbucks didn’t stop there. In 2015, the company started to offer a way to pre-order drinks and pay in-store, which allowed customers to skip lines and pick up their drinks in any store in the US. This dramatically increased the number of transactions going through the app, providing Starbucks with a wealth of valuable information about users’ spending habits across the country. In 2017, the app has even become the top US mobile payment platform, ahead of Apple Pay and Google Pay!
Using data to create value
An interesting metric to understand how Starbucks thinks about consumers would be the customer lifetime value, which has three main components: average purchase price per customer per visit, the number of visits per customer per year, and the average customer lifetime. Starbucks realized that it could maximize all three of these components with data analytics, making it a more powerful competitive advantage than the quality of its coffee (because, let’s be honest, we’ve had better lattes elsewhere). Let’s dive into a couple of ways the company uses data analytics to increase customer lifetime value.
First, data has helped Starbucks by personalizing the customer experience. By analyzing consumers’ historical orders and purchase patterns, the app can not only suggest customized recommendations but also push tailor-made offers that are likely to increase foot traffic on any given day. For instance, if you usually buy an espresso on Tuesdays but did not come last week, you can safely expect a notification inviting you to purchase a drink for twice the amount of ‘stars’ you usually get. This directly impacts the average number of visits per customer. Alternatively, the app can also invite you to try a new product based on your preferences, which will probably increase your spending per visit on this specific day.
Secondly, a more behind-the-scenes way that Starbucks leverages its enormous amount of data is by using it to offer brand new products. Innovations like non-dairy or unsweetened alternatives, summer special drinks, or new home consumption products were born thanks to the scanning of users’ preferences and extrapolation of such data.
Last but not least, one of the best levers to increase the average customer lifetime is to make it easy for consumers to keep coming to Starbucks, whenever they feel like having a coffee break. In other words, there should be a Starbucks store everywhere there is potential demand. So the company also uses its data to plan where to open new stores. Artificial intelligence makes revenue projections based on factors such as income levels, traffic, or competitor presence, and helps determine where the next big revenue opportunity is.
Starbucks thereby has all the main characteristics of a data company, the biggest one being that it is self-sustaining: the more data Starbucks collects, the more it can make the right decisions to grow its business, thus growing the number of transactions which in turn helps make even better decisions.
A key question is: what more can Starbucks do? How can they leverage data and machine learning even better in the future? I can easily see Starbucks’ moving even further and automating the purchase experience, the biggest bottleneck to growth being the human factor. The only problem with that is that the company would then get rid of its initial success ingredient: a smiling, customer-friendly team. Even though people might want a faster, convenient experience, who would say no to a barista handing a warm cup with your handwritten name on it, wishing you a good rest of the day?
The Verge. 2021. Starbucks says nearly a quarter of all US retail orders are placed from a phone. [online] Available at: <https://www.theverge.com/2020/10/30/21540908/starbucks-app-q4-earnings-mobile-payments> [Accessed 23 March 2021].
The Verge. 2021. Starbucks rolls out mobile ordering to all US locations. [online] Available at: <https://www.theverge.com/2015/9/22/9366617/starbucks-mobile-ordering-now-available-everywhere> [Accessed 23 March 2021].
Medium. 2021. Starbucks Isn’t a Coffee Business — It’s a Data Tech Company. [online] Available at: <https://marker.medium.com/starbucks-isnt-a-coffee-company-its-a-data-technology-business-ddd9b397d83e> [Accessed 23 March 2021].
Student comments on Starbucks: From Coffee Machines to Machine Learning
This is really insightful Stephane! I had no idea that the Starbucks app was the top US mobile payment platform and I agree that it is a major differentiating factor compared to other coffee shops. I’m sure that the app also speeds up transaction time and enables Starbucks to process more customers in a given time period, thereby increasing revenue and keeping customers happy since they don’t have to wait in line as long. One question that came to mind is whether the company has been able to A/B test in-store features by using the Starbucks app? We’ve seen many examples of A/B testing on digital platforms but I would love to learn more about whether Starbucks A/B tests things such as specials, types of ingredients, physical ambience? I think that the Starbucks app could be an example for other retailers of how to collect data and use it to A/B test non-digital aspects of products and business models.
Thanks for your comment, Tiffany. That’s an interesting point, and surprisingly I could not find any concrete evidence of Starbucks using A/B tests to improve the user experience. I can however definitely see how this can be tricky in a non-digital context, such as measuring the store experience. In a physical environment, there are so many factors to control for (location, store size, popularity…) that I wonder if it is even possible to conduct a proper A/B test as you would in an exclusively web-based experience.
Super interesting. To your question of “what else can they do” I’ve always wondered if there was an opportunity for Starbucks to white label their digital payment and app technology. Basically becoming the AWS of quick service restaurants. It could start through providing just a simple front-end, but your post shows the far reaching impact they could have across the business.