Ratings, reinvented: Bundle's business ratings can tell you things no other rating site can
So, Alex, I know you, but why don't you go ahead and introduce yourself.
I'm Alex Hasha, and I'm the chief data scientist here at Bundle, and that basically means I'm responsible for transforming a huge database of credit card transactions into a business review site that people can use to find cool places to go, and make better decisions with their money. We're trying to build a website like Yelp, but it's entirely driven by data rather than user reviews.
Like a lot of people, I use Yelp. I use it to get a general sense of a place I haven't been to, and figure out if it's good and worth my time and money. How is Bundle different?
I think Bundle and Yelp are different even if they fall within the same category of websites. You kind of have to unpack the question of what it is you're looking for when you're looking at a business listing site. There's a lot of questions you might be asking, but the main question you have is probably: "What's a good place?"
But, how do you define "good"? At Yelp, "good" is based on the opinions of people who have been to a place — or claimed to have been — and have gone onto Yelp's website to post a personal review. Yelp is very strong at collecting personal descriptions of what a place is like. A person can tell you about the ambience and what their personal experience was, and that can be useful. Even so, there's a lot of system gaming on Yelp that can mislead you into believing that a good place is bad, or a bad place is good. For example, an owner of a restaurant may enlist friends and family members to post overly positive, biased reviews to make a restaurant sound popular even if it isn't. Similarly, competitors can place nasty reviews on another business's Yelp page to provide a negatively distorted picture of what a place is like. We've gotten used to navigating that, and putting in an effort to make sense of what we're reading, but it doesn't have to be that way.
A search for the popular Back Country BBQ in Dallas, Texas instantly tells us that it is a popular destination (a five-star Bundle rating), and that based on a sample of more than 500 transactions, diners typically spend about $20 when they go.
So what can Bundle tell me that Yelp can't?
Our reviews are based on billions of credit card transactions from customers of different businesses, rather than a random person's personal review. Because our data is based on anonymized credit card transactions collected from tens of millions of real credit card holders, we have lots of transactions from almost every business that accepts credit cards, and that includes all the popular places, and the less well-known ones too.
We're also not prone to rampant meanness, or the system gaming that happens on Yelp because our reviews are generated by people who don't know that they're generating reviews. They're going about their lives and creating data by using their credit card. Obviously, we anonymize their data so they can't be linked to their transactions, but we see that many individuals have been to a place, and how many people have returned, and how much they tend to spend when they go. These are all things that create a very helpful picture for someone who wants to learn about a business they've never been to.
But since Bundle uses credit card data, that means you can't rate places that are cash-only, right?
Right. And we use just a sample of the vast amount of credit card transactions out there. But we still have billions of credit card transactions to work with, and fortunately, most businesses accept credit cards.
How are you making sense of this data?
You can think of credit card data as a big survey of people's tastes. Everybody gets a certain number of dollars, which you can think of as votes, and then they have to spread them around. They have a limited number to use, and they have to make choices to where to use them, and in that sense you can think of a dollar spent as a vote in favor of a business. And you can sense that a place is good if it's getting a lot of people's votes.
Tell me about the Bundle rating and how it works.
We started thinking about what we could measure just by knowing that someone swiped a card. And the big thing that we thought was important was repeat customers. If we see a business has 3,000 customers, that might make them popular, but if none of these customers ever come back, that detracts from your rating. It might be because you're not in that part of town often, but it could be simply that you didn't like your experience, so you didn't go back.
We also want to take a look at the average prices people are willing to pay. Because it's easy to get a lot of people to come to your store by charging low prices, but it says a lot more about how much a person likes a place if you can charge high prices and get repeat customers. So higher prices people are paying will increase your rating as long as your popularity goes up with it.
In short, the Bundle rating is based on popularity (or how many people go), repeat visits (or how many people come back), and amount spent per visit (how much people are willing to pay).
How often are the ratings updated? People can be fickle. Popular places can quickly become less in demand due to the time of year, or because customers' preferences have changed.
Our data gets updated on a weekly basis, so Bundle ratings would change periodically after the updates are made.
Back Country BBQ's rating of 93 places it in the top ten percent of restaurants in its area, a good indicator that people enjoy dining there. Ribs, anyone?
The rating ranges from 1 to 100. What's an average rating, and what's a good rating?
We set it up so the average business gets a 50. So a 75 means you're in the top quarter, which is pretty good. And a 90 means you're in the top 10 percent, which is really good. There's nothing wrong with a 50 or a 40. That just means it's an average place, or just below average. Most places out there are average.
In addition, each business's rating is relative to other businesses in the same category in the same city. And we have dozens, if not hundreds of categories. In other words, a restaurant in Dallas is going to be compared to other restaurants in Dallas, and not a clothing store in Dallas, or a restaurant in New York. We try our best to compare apples to apples to account for customer preference between a collection of similar options.
So, right off the bat, if you see a rating that's above 80, you can pretty much tell that it's going to be a good place compared to other places in the city you're searching in?
We find that if we start actually using the rating to make decisions that, more often than not, it leads you in the right direction. With Bundle, I like to use the ratings to narrow my choices down to a set of places that meet my needs that have high ratings, and then, right on our website we have links to Yelp, Citysearch, and other reviews so I can see a little more detail.
Business listings on Bundle also have links to reviews on other sites to help give you a complete picture of what a place is like. Above: Related reviews for Le Bernadin, a renowned New York restaurant.
What are the other things Bundle can offer?
We have a lot of great features in progress that we're constantly expanding and improving. We can tell you a typical cost people pay when visiting a business, which is a little different from what you'll see on other websites, but also, more informative. We show the median transaction size of people who've visited a business, which includes individuals, and parties of two or more. Just like on Yelp, where you have to look at reviews and judge what they mean to you, on Bundle you have to look at prices and judge what they mean to you. We show the range of prices people have paid. So if you're planning on figuring out what a single individual would pay at a restaurant, you'll want to look at they typical costs below the median. If you want to see how much a large group would pay, you would look at what people are paying above the median.
We also break customers down into groups based on customer behavior. So we have segments like high-rollers — people who we've seen love luxury brands like Burberry, Prada and Gucci — and we can see what kind of restaurants and grocery stores they like to eat and shop at. We other segments — foodies, for example, and locals. So these are other dimensions we can provide to help you get a better sense of what you're looking for.
Give me an example of how you've used the Bundle rating in your own daily decision-making.
A friend of mine moved to Hong Kong a few years ago, and every time she comes back to visit she wants to have a big dinner party with a bunch of friends. And the last two or three times, we went to this restaurant in the West Village in New York City called La Ripaille. It was pretty enjoyable, but both times I felt like it was too expensive for what it was.
But I had really enjoyed the experience because it was really French-y, and the owner would come over to the table and tell us in this thick French accent what the best things in the menu were. I loved this. So I wanted to find another French restaurant like this, but where I didn't have to pay as much. So I looked up La Ripaille on Bundle, saw that the Bundle rating was 73 — above average — but the typical cost was about $121, which was actually close to my own experience. I had paid for myself and my wife, and at the end of the night, I felt like I didn't have that much to eat, but I had paid that much. Fortunately, we have another feature on our site where we can show what other restaurants people who visited La Ripaille also visted:
So, I looked at that feature to find restaurants customers of La Ripaille had also been to and liked a lot, and one of them was a restaurant called Cafe Cluny, which was also French, but instead of a rating of 73, it had a rating of 93. Its typical cost was $58 instead of $121, and when I expanded the "Who Goes Here?" tab, I saw that the restaurant was popular among foodies and locals. Finally, I looked at the "Reviews" tab and saw that there were tons of reviews and people were very positive about it, so I thought, "okay, this is where I'm going to go." And we went and it was great — the food was great, and I ended up paying about $70.
So, the Bundle rating turned out to be true to life. Final thoughts about the Bundle rating?
I think the key role that Bundle played was helping me find what I was looking for so much faster. Because there's no easy way to say, "I want to go somewhere like La Ripaille, but cheaper and better. But on Bundle, I was able to do that in under a minute. And this is just the beginning of what we're trying to do to help people make better spending decisions.
Find the best places near you. Click here to get started.