People are keen to find a good place to eat. The proliferation of surveys and sites like Michelin, Zagat, The Infatuation, and the plain ol’ newspaper restaurant review are signs that a professional opinion still matters — especially when South Park serves Yelp so hard. Such is the thinking of Renzell, a new data and media company that’s employing its own ratings methodology to rate fine-dining restaurants — and to rate the survey-takers, themselves.
Bo Peabody, Renzell’s founder, spoke to me from his offices in New York. Peabody wholeheartedly believes Renzell can be a better ratings company, and that its algorithms can be used across the world. He’s so committed to transparency — Renzell shares its findings with the restaurants themselves — that he sent me a log-in to go through a survey. Like Zagat, Renzell depends on diners to fill out forms on their experiences. Peabody sees a lot of differences between his troops and Zagat’s. Renzell’s survey itself is impressive, and I’ve seen a lot of Zagat paperwork in my time. Cleanly built, it’s focused on all manner of the dining experience — from the soundtrack to the pacing of the meal. It was fun to fill out. Peabody’s banking on that.
Inverse: Can you give me some background on the company?
Peabody: The last 20 years I’ve spent living two parallel lives: One as digital media technology venture capitalist and then the other as a restaurateur. I’ve had two restaurants. Renzell is, for me, a culmination of a lot of work that I’ve been doing. So, living in Manhattan for the last 15 years — as someone who loves restaurants and who is in the industry — I continue to eat out a lot. I was always struck by how antiquated the sort of ratings and reviews ecosystem is, not only in New York, but all over the world. In New York, technological innovations have impacted virtually every other aspect of life.
It also struck me that if you can collect more data-driven reviews, you can also give that data back to restaurants to allow them to continue to improve the guest experience. I guess the other observation I made is that these places — whether it’s Michelin or Zagat — come up with ratings and then put them in incredibly ugly phone books. The subject they’re covering is one of extraordinary beauty. So, I set out to solve those things: Let’s come up with a more data-driven approach to creating ratings that allows you to eliminate a lot of the amateurs that plague all the other systems. Then, let’s share that data back with the restaurants, so the whole thing doesn’t feel so opaque and weird. Then, let’s put the ratings in something beautiful that is consistent with the beauty of the subject matter we cover.
Okay, and you’re doing that once a year?
At least in a public way, we will be issuing the ratings once per year. They will appear in the first issue of what will be a quarterly magazine. Then, the following three issues will have other interesting data tidbits, but also just beautiful stories and features about the restaurants that we cover.
I had to come up with a way to limit the total number of restaurants to a definable universe. We also decided to approach that from a data-driven perspective. We began to track — about a year-and-a-half ago — basically all the restaurants in New York City that would be considered. We started with 225 and the list has grown to 265. We track all of those on 32 different characteristics and we give each restaurant a score on those characteristics. It’s a simple algorithm that we weigh the 32 things in a particular way — some are more important than others — and that’s how we came up with the original list of who to cover.
Did you use data from other sources to get there?
Yeah. About half of the things that we have are existing data, like Wine Spectator, Michelin, and then the others are primary research that we did on our own. So we’re not really saying “Michelin is bad.” I just think it is a singular concept; it is flawed.
Probably the thing that is most unique about our business is once we choose those restaurants, then we do a very deep data methodology on getting at the experience that guests are having over multiple nights over the course of multiple experiences with multiple profiles of people.
Can you tell me more about the data itself?
We’re sharing it with everybody. Most of it is on the website. We’re being totally transparent with the restaurants about the data we collect. Michelin doesn’t really have data. Michelin and Zagat exist on the opposite ends of the problem spectrum: Michelin is plagued by gross subjectivity. They only have three or four people in each city that are eating at these restaurants. They’re only eating at these restaurants three, four, maybe five times. So you have a very small amount of people, who have their own biases. At the other end of the spectrum, you have the opposite problem where you have too many people, most of them have no idea and no business opining on the high-end restaurants. I think their opinions on the place on the corner are probably fine, but the fact is — for better or for worse — there’s a small group of people who are really qualified to talk about all the aspects that a great restaurant should have.
What we set out to do is something in the middle, where we have a curated group of people that will be somewhere between 500 and 750 people in each city. We started off with 75 people from our personal networks, the six of us that started working on the project. I interviewed 40 of them for an hour to make sure they knew what the heck they were talking about and, then, 38 of them passed the test and we invited those 38 people, and then we invite the other 35 somewhat blindly. We started off with about 65 people in a beta test in May. Once we saw what we they were doing, we allowed them to start referring other people. When somebody gets referred, we then do our own research and we build a profile of who these people are. People will apply, we take them through an application process. But people who are invited by existing members, we do a bunch of research on them. The real thing that’s important is that we’re using the data that we’re collecting on these people and we’re weighting those answers to the surveys based on those things that we know about them. So, in the world of technology companies, this is rudimentary data science. But, in the world of restaurant ratings, this is revolutionary.
Were there any restaurants that were unexpectedly rated highly?
Absolutely. I’ll give you an example. There’s a restaurant called Taboon in Hell’s Kitchen and the chef basically brought high-end Middle Eastern cuisine to a fine-dining format. He left and just recently returned, and it’s sort of considered a neighborhood restaurant, but our data suggests that on every dimension, it’s a destination restaurant.
I don’t see Per Se.
Yeah, that’s the other end of the spectrum. I’d say the places that people are most often surprised by not being on there are Masa and Per Se. You know, the data suggests that — you can see what we track — and those places literally score zero on value. They score zero on vibe. And when you get zeroes in categories, it’s very difficult. The way we weigh the categories, those restaurants are never going to do well. We are editorializing. I’m not washing my hands, but we definitely have a point of view about which of those 32 things are more important.
I can tell you that Michelin star is not one of the things we weighted very highly. It’s not that we don’t respect Michelin, it’s that Michelin is focused on a very particular point of view and there’s a very limited set of things that we don’t think are very indicative of what you’re eating and caring about. If you go to Per Se, you’re going to sit with a bunch of tourists. And it’s in a mall! I mean, look, I’ve been there several times. There are things about it that are fantastic. I think it’s something everybody should do if they can afford it, but I don’t think it’s a place that you’re going to make a point to go to every year.
Hello! You've made it to the end of the article. Nice. Here's a related video you might like: "The Netflix Algorithm Is Killing Genre Bias, New Data Shows"