Grocery and recipe delivery service Hungryroot has a revolutionary method for collecting first-party data from its subscribers.
Wait for it … it asks them directly.
People are willing to volunteer information about themselves if they know how it’ll be used, if there’s a clear personal benefit and if they trust who they’re sharing it with, said Alex Weinstein, chief digital officer.
“Online data collection is so pervasive, and people are so used to being tracked online in the name of being shown ‘better ads’ that it makes them wary of sharing their data,” Weinstein said. “But if we explicitly ask someone, ‘Are you allergic to anything?’ the only reason we want to know is so we don’t include that ingredient in their box.”
What’s cooking?
Hungryroot is a little different from other meal delivery services in that it makes personalized grocery recommendations that can be used for meal prep, as opposed to delivering preprepared meals or recipe cards with ingredients, like HelloFresh or Blue Apron.
As part of its onboarding process, Hungryroot takes users through a lengthy quiz about their dietary needs with questions on everything from how many people they need to feed and what their goals are (losing weight, for example, or getting healthier) to how much time they typically spend cooking and which kitchen appliances they use most.
Hungryroot also asks about dietary restrictions, favorite cuisines and flavor profiles, preferred proteins, which foods people dislike, desired spice level and favored snacks.
It’s a lot of questions, and conventional wisdom states that a sign-up funnel should have as few steps as possible. But Hungryroot has found that friction works in its favor. People don’t mind being asked a lot of questions – and conversions actually improve – as long as those questions are relevant to the process, Weinstein said.
“The promise here is that we’ll really get to know our customers and populate their cart with personalized foods they’ll like,” he said. “So the more questions we ask, the more people believe that we can deliver on the promise.”
The celery question
But Hungryroot also continues to gather more nuanced information about customer likes and dislikes after their personalized grocery deliveries begin to arrive.
And this involves a lot more than getting a thumbs up or down.
If someone tells you they don’t like salmon, then you don’t send them salmon. That’s table stakes, Weinstein said. But learning about customer preferences gets a lot more complicated when it comes to recipes that have multiple ingredients.
Which is where AI plays a role.
“If someone says they didn’t like our chicken salad recipe, does that mean the person never wants another chicken salad recipe from us, or is it that they didn’t like one of the ingredients, like celery?” Weinstein said. “Knowing the difference depends on us understanding why you said you didn’t like the chicken salad.”
Not knowing the “why” can mean the difference between churn and retaining a customer.
Say Hungryroot doesn’t detect that celery was the culprit and sends multiple recipe suggestions with celery as an ingredient. Eventually, the recipient might get fed up and leave.
“The power of AI is about being able to draw inferences from the data we already have and make good recommendations,” Weinstein said. “That’s how you move forward in your understanding of the customer.”
Hungry for insights
Hungryroot uses some open source AI software, but most of its technology is proprietary and was built in house. Weinstein has a digital team of around 55 people, including data scientists and ecommerce experts.
So far, they’ve uncovered some interesting insights from first-party customer data that they’re able to apply more broadly to the company’s personalization strategy.
For instance, Hungryroot needs to account for flexitarians who primarily maintain a vegetarian diet but occasionally eat fish or meat. It wouldn’t make sense to start recommending burgers every week to a flexitarian who only wants beef occasionally, but it would make sense to surface beef ingredients every so often.
Hungryroot also has to consider how people define certain terms.
Being vegan, for example, means something very specific. If a customer tells Hungryroot they’re vegan, the system automatically suppresses anything that doesn’t conform to that diet.
But Hungryroot has noticed a trend whereby some users consider themselves to be vegan, yet want to add certain milk-based products to their cart.
“We need to be thoughtful about exactly how we ask people about their preferences because we want to make sure we’re being helpful without being invasive or making someone feel bad about their choices,” Weinstein said. “This is something we’re experimenting with.”
Rejecting the RMN model
With so much first-party data about its subscribers, it would seem only natural that Hungryroot hop on the bandwagon and launch a retail media network (RMN) of its own, similar to grocery delivery service Instacart.
But the company has no such plans, Weinstein said, and there are no traditional display ads anywhere on Hungryroot’s website.
It makes sense why others have embraced the RMN model, Weinstein said.
Paying someone to walk up and down aisles in a busy grocery store, pick items and then deliver them to the end customer is a very expensive, relatively low-margin proposition, he said. Which is why many of these services are turning to high-margin online advertising.
“But we’re operating on a 40% gross margin, we’re growing, and we’re focused on differentiating through personalization,” Weinstein said. “We’re not in a position where we’re being forced to diversify into a different business.”
