Food is emotional. That simple reality sits at the center of the data strategy of food delivery app DoorDash.
“What one person likes, another person really doesn’t,” said Jessica Lachs, who heads analytics for the company, at the Transform conference Wednesday in Mill Valley, California.
It would make sense, then, that showing DoorDash users pictures of food would whet their appetites more than showing them pictures of restaurant logos — and that’s what DoorDash data analysts expected to find in data from an experiment.
Spoiler: They didn’t get the results they expected. The mishap illustrated a caveat about data in retail. In a sector where businesses rely heavily on the amorphous it-factor of customers’ emotional connection with food or merchandise, data can seem to give answers that are distant from reality.
June 5th: The AI Audit in NYC
Join us next week in NYC to engage with top executive leaders, delving into strategies for auditing AI models to ensure fairness, optimal performance, and ethical compliance across diverse organizations. Secure your attendance for this exclusive invite-only event.
When DoorDash swapped boring restaurant logos for images of actual food in its app, “what we initially saw was conversion dropped,” said Lachs.
She added, “That kind of happens a lot where you go into something expecting a result.”
The actual problem that caused the dip in conversion turned out to be counter-intuitive.
“What we had found was the algorithm was choosing the most popular item,” said Lachs. “That makes sense. However, at a lot of merchants, the most popular item isn’t necessarily the most representative of what the merchant offers.”
An image of chicken wings popped up for a pizza place. A sad-looking salad represented an Italian restaurant.
DoorDash changed the algorithm to pick food that was more representative, and the company saw the bump its data analysts had expected. The original hypothesis was right.
DoorDash isn’t alone as a retail company to find that data can’t act as a crystal ball. Heads of data teams at Nike and Target shared similar views.
“We’re good at figuring out what works, how people are feeling directly about it,” said Timothy Sears, vice president of data science and engineering for Target, adding, “Getting that introspection out of [data] — not so much.”
Nike tries not to let data shape customer experience into something “too robotic,” said the brand’s vice president of data and analytics, Gus Weber.
“The emotional connection is what makes retail interesting,” he said.
None of that is to say data can’t inform better decisions for retailers. The bottom line is companies in the space need to use the right data for the right purposes.
DoorDash has had success using data to customize the app for different users. If you order the same meal from the same restaurant at the same time on the same day every week, the app might send a reminder saying it’s time to place your weekly order. If you order from all different types of restaurants at all different times, the app might show you new and varied restaurants.
And Nike has used customer data to better understand the type of customer the brand is interacting with. Is this person enticed by collector items, or are they just looking for a new pair of running shoes?
The aim with data, said Weber, is to figure out “how we think about, ‘What is your intent with us?'”
