"When you recommend a restaurant dish, you're asking people to commit time and money to actually trying it. So there's a risk they will get mad at you if they don't like it. We get that. That's why, instead of telling users to eat one thing from one restaurant, our Foodspotting app gives them options, and we're transparent about our process. We explain, very clearly, why Foodspotting is showing each food to them--maybe a critic they like recommended the dish, maybe the flavor profile is similar to another dish they often eat--but what to eat is totally up to them. Eventually, we may be able to offer up single recommendations. But you have to wonder if people would even trust us. Maybe, with food especially, they like to have choices."
Fast Company: Where did you get the idea for Foodspotting?
Alexa Andrzejewski: I was inspired by traveling to Japan and learning about how many interesting dishes there are out there that I didn't even know existed. I came back and I wanted to share these interesting finds with people and also find them locally, but I realized there was no way to use existing apps to search for a specific dish. So if I was looking for Japanese curry, or even a milkshake, there was no way to find the best milkshake in San Francisco. So it just seemed like such an obvious opportunity.
What's tricky about recommending food?
Food preferences are something very personal. I don't even like the same foods that the rest of the Foodspotting team likes, even though we work together and see each other the most. I think a lot of apps focus on social recommendations above everything else; with food, social is just a part of it, so it's really important to not weigh social signals too highly, rather than finding out which people actually have the same taste as you.
So who else do people trust?
Expert recommendations have a lot more sway than you realize. In San Francisco, 7x7 publishes this list of "100 Things To Try Before You Die," which has been incredibly influential on the dishes that everybody [in San Francisco] knows and recommends. We realized that's an important decision-making factor, knowing that something is Michelin-recommended or Zagat-rated. So we try to surface that information as well. When it comes to deciding what to eat, people prioritize personal taste, celebrities, and then maybe friends.
How do recommendations from an algorithm play into that trust?
One of the big things our users have said is that they don't necessarily trust recommendations from algorithms; they trust word-of-mouth. And that was an interesting insight: that word-of-mouth is how most people make decisions about eating day to day. It's not by checking an app--it's about some shaved ice place that everyone talks about. We're trying to replicate that experience of word-of-mouth on Foodspotting. We're not trying to be "magically smart" with our recommendations--we're trying to use what we know about people to find good matches for them. I love ramen, for example, and I'm going to be impressed when Foodspotting says, "You like ramen? Here's great ramen," even though I know it's doing something relatively simple.