A Data Scientist's Guide to Finding Good Restaurants

How to avoid being overwhelmed by the "big data" of thousands of restaurant reviews, and better your chances of stumbling on a restaurant you'll love

My wife and I love trying new restaurants. We also love to travel – but the main thing we focus our travel itinerary planning on is where we’ll eat. We are fortunate to live in a time when social media and online reviews make it easier than ever to learn about a restaurant long before we come close to darkening its doorway. It’s a bit staggering that today it’s possible to plan a detailed food tour from the comfort of your home for a town thousands of miles away that you’ve never set foot in before.

But of course, finding restaurants you’ll like isn’t so easy, even when doing this research is easier than ever before in human history. I’d say we strike out far more often than we hit a home run, despite our best efforts. However, I’d like to think we’ve also learned from our mistakes: while our success rate is far from a perfect 100%, it’s also a lot better than what it used to be.

I’ve had a lot of occasion to refine my analysis methods here over the past few years – between vacations, working remotely while traveling, and taking a nearly year-long sabbatical from work, we’ve traveled to plenty of new places where we’ve had to find a place to eat while being completely unfamiliar with the local restaurant scene. A year and a half ago we moved to Kuala Lumpur, the city where I grew up, and we’ve had quite a journey navigating the food scene both here and across the rest of Southeast Asia. So be forewarned: this will be a long post.

The tips and tricks here don’t guarantee you’ll be successful most of the time – most restaurants we try are still somewhere between just ok and disappointing. And while I’ll try to make my advice as broadly generalizable as possible, it’s of course all my own opinion – and especially when it comes to food, so much of this is subjective and up to one’s taste. I don’t intend to yuck anyone’s yum here – I’m just sharing what I personally find helpful. You might want to modify my advice, or disregard it altogether – and with how personal food is, that’s more than fine!

Finally, before we begin, don’t worry – I may be a data scientist, but none of my advice here asks you to do any number-crunching beyond what online review aggregators should already be doing for you. Most of the analysis I do is qualitative, and when it comes to the quantitative, my advice today will be more about how to interpret numbers than anything to do with actually doing something with or to numbers.

In short, here’s my approach to restaurant research:

  1. Understand your own priorities – what kind of restaurant are you looking for?

  2. Find restaurant recommendations from information sources (friends, bloggers, creators) that are likely to share your priorities and preferences

  3. Make a list of candidate restaurants based on these recommendations and add any others that seem similar to what they recommend (often restaurants in the same neighborhood will also be worth considering)

  4. Scan the reviews for red flags, and eliminate restaurants accordingly

  5. Make your final selection from this shortlist

If any of this sounds interesting to you, let’s dive on in.

Know Thyself

The first thing you need to do is to be honest with yourself about what you’re looking for. As a professional analyst, you’d be a fool to start any project without first knowing why the business is asking for this analysis. Finding a place to eat is no different: what’s important to you, and the people you’re eating with, about a restaurant?

There are many possible things you might prioritize, and how you prioritize them is not only intensely personal, but also very contextual – there is not a one-size-fits-all-the-time answer here. Here are a few things that might matter to you:

  • Taste – how important is it to you that you enjoy eating the food itself? Obviously it’s always imperative that the food be edible – but beyond that, different people can place very different levels of importance on the taste of their food.

  • Atmosphere – what kind of space are you looking to eat in? Somewhere quiet, or somewhere energetic? Somewhere fancy, or somewhere homey?

  • Location – how far are you willing to travel? How important is it that the restaurant be close to where you’re coming from, or where you’re going to next?

  • Genre – do you want to try a new cuisine, or eat something familiar? Are you looking for something special, perhaps a once-in-a-lifetime experience, or are you looking for something very familiar and comfortable as part of your daily routine?

  • Cost or value – are you working within a particular budget? Is a meal that’s fantastic value for money a reward in of itself for you?

  • Service and Accommodations – do you have dietary restrictions? Are you bringing a large group that you want to be sure the restaurant can seat and serve quickly? Do you want a waiter who will be friendly and make conversation with you? Are you looking for special white glove service that makes the occasion feel special even if everything else is just mid?

Your priorities may be quite simple and rarely dynamic – my wife and I virtually always prioritize taste over all else, and are flexible on most other dimensions other than location, which typically is our most binding constraint when picking a restaurant. But despite this, we often spend a lot of time discussing what type of cuisine we’re feeling like – this shifts day to day, if not hour to hour. And while we’re often just planning day-to-day meals for the two of us, our priorities obviously shift when we’re planning a meal for a group, or for a date night. Circumstances often dictate that our priorities need to shift.

A plate of tuna tostadas

Some of the best tuna tostadas I’ve ever had, at Entremar in Mexico City

Knowing what you’re looking for may seem basic, but this is so critical to everything else I’ll talk about today. If you aren’t sure what you’re looking for in a restaurant, all the 5-star reviews in the world are useless to you. It’s easy to say you want to eat at the best restaurant – so why aren’t you eating at a 3-star Michelin restaurant every time you dine out? Obviously, cost is one reason – but the truth is, hardly anyone would actually enjoy having most of their meals in a fine dining restaurant, no matter how amazing the food, atmosphere, or service might be. At the other end of the spectrum, Southeast Asian street / hawker food is delicious and cheap – but my wife and I can’t eat it for every meal, because we have other things we’re optimizing for than just taste and cost.

We are all optimizing for different things in a restaurant, and unless you know what you’re looking for today, you won’t be able to effectively narrow your search or interpret the reviews and media you see about the restaurants you’re considering. Every reviewer is optimizing for different things. Some reviews you read or watch will be from people who care about things that don’t matter to you; you should ignore them. And when you find a review from someone you’re confident is optimizing for the same things as you, it’s as good as gold – that one datapoint is worth far more than hundreds of reviews from people who don’t share your priorities.

Find Your Cohort’s Recommendations

This brings us to our next point: once you know what you’re looking for, find the people who think the same way as you. This means people who prioritize the same thing in a restaurant, and also as much as possible, people with the same tastes. Even if two people care a lot about finding a restaurant with a cosy atmosphere, what’s cosy to one person can be cramped and overcrowded to another. If you’re looking for yummy food, you want advice from reviewers who share your tastes in what’s yummy or not!

This all sounds very theoretical, so let’s translate this into some practical advice by way of example: if you are visiting a new city and your number one priority is finding yummy food near where you’re staying, it’s often best to ignore all the highly-rated restaurants on Google Maps near your hotel. The more reviews and the higher their star rating, the more confident you can be that you won’t have a good time. There are many reasons why, but a big one is this: most tourists don’t care as much as you do about finding the tastiest food, so the restaurants near where the tourists stay won’t be serving it. For a lot of tourists, the priority is finding somewhere to fill the stomach that’s safe, comfortable, and convenient so they can fuel up for a day of travel. Restaurants that can do this close to the tourist quarter will consistently earn good reviews from most of their customers, because this is what most tourists are optimizing for.

A screenshot of Google Maps search results for the best restaurants in Barcelona

The “best restaurants in Barcelona” all have thousands of reviews – a potential red flag

Review aggregators like Google Maps can be extremely useful tools, but you need to use them prudently. The overall star rating and number of reviews are just not that useful in areas full of tourists, unless your priorities are that of the average tourist. To find restaurants worth considering as a tourist whose top priority is delicious food, you should start your search with information sources that are dominated more by people who share your priorities.

Because it’s so hard to tell whether the average internet reviewer has my priorities in mind, I’ve found it’s best to start a restaurant search by looking at blogs and social media from people who have given some signal that they fit my profile. For example, I’ll try to find bloggers and influencers whose content primarily focuses on food, rather than travel. I’ll ask for recommendations from friends who know my tastes, or look to creators I’ve come across in the past whose content suggests they like things I enjoy as well.

Don’t get me wrong: I often start a restaurant search by getting oriented with the area I’m going to on Yelp or Google Maps. But I typically avoid starting with a “data-driven approach” that just looks for the best-rated restaurants on a review platform. As long as they aren’t 3 stars or below, I won’t pay attention to which restaurants have the best or most number of reviews.

To generate a list of restaurants worth trying, I’ve found it’s often easiest to find a blog or vlog (on YouTube/TikTok) from someone who’s likely to be of the same mindset as you. If you’re vegetarian, find creators who talk about looking for vegetarian food when they travel. If you enjoy adventurous eating when you travel, watch The Best Ever Food Review Show (seriously). Do this a couple times and you can assemble a decent list of candidate restaurants – you’ll notice the same restaurant names cropping up from different creators. Then you can look on Yelp or Google Maps and see what other restaurants are similar to the ones you initially shortlisted – highly-rated restaurants in the same area are often worth considering. This doesn’t get you a final decision about where to go for dinner, but it at least gives you a few restaurants to choose from, and then you can apply the other tips below to further narrow down your choice.

Parse the Reviews

Once you’ve compiled a shortlist of restaurants to try, then we can actually start looking at some data and look at reviews from aggregators like Google Maps or Yelp. In North America, you should start your restaurant search on Yelp rather than Google Maps or TripAdvisor. Yelp got its start as a restaurant review platform, and so I’ve found its average reviewer is much likelier to be a foodie than the average reviewer on other platforms.

I learned this the hard way when I once booked us at a resort in Mexico that had good reviews for its food on Google and TripAdvisor. We found most of the food this place served close to inedible, and couldn’t believe anyone would have rated this hotel highly for its food. I decided to look up the hotel on Yelp, and true enough, both the overall star rating and the reviews themselves would have been a dead giveaway that the food here was not for us. The restaurants at this resort were perfectly fine for someone prioritizing other things on their vacation – but the Yelp reviewers’ take matched ours, because our priorities matched theirs.

If you’re outside the US, Yelp is most likely going to be irrelevant to your search – I recommend Google Maps if so. TripAdvisor is a famously manipulable platform (see the guy who made his shed the #1 restaurant in London on TripAdvisor), but its reviewers also tend to be tourists, which automatically makes them a less reliable cohort of reviewers unless your preferences align with the median tourist’s. Google Maps and Yelp aren’t perfect but they have a much better chance of getting reviews from locals and non-tourists than TripAdvisor does.

Whether on Google Maps or Yelp, if you’re like most people you will probably skim the photos first – does the food look like something you might enjoy, or at least be curious about? Does the ambience of the place look like somewhere you’d be comfortable dining in? You can pretty quickly discard restaurants that don’t look right based on this. But the text of the reviews themselves also present a good source of data for ruling out certain restaurants – so let’s talk about how to skim reviews efficiently.

I’ll primarily focus on ruling out restaurants because from experience it’s hard to find a restaurant you can be confident you’ll love from research alone – it’s just far too personal for you to know until you’ve been there for yourself. The advice here will mostly cover red and yellow flags you can see from a quick skim of the reviews that’ll help you narrow down your choices.

The first thing I do with a restaurant I’m considering is scan the top-ranked reviews – these are often ranked based on how “helpful” Google or Yelp thinks they’ll be, and so I use them to get a rough idea of what the restaurant is like. If a bunch of reviewers consistently mention something that’s a big turn-off for me, we can rule the restaurant out pretty quickly. 

The other thing I look for from a quick skim is evidence of fake or purchased reviews – an unfortunately common thing in Southeast Asia, but also many other touristy areas I’ve been to across the world. One easy giveaway is if most of the top reviews are brief and don’t give much detail. A less obvious giveaway is when almost every review mentions the waiter by name:

A screenshot of reviews from Google Maps for a restaurant in Hanoi where each review mentions their host by name

This Hanoi restaurant has pages upon pages of 5-star reviews all mentioning their server by name

One or two reviews mentioning a particular server are normal. If it’s almost every single review, then the restaurant likely has a policy of having the waitstaff ask customers to leave them a review, and to mention which person served them. (I imagine waiters might be compensated based on how often their name appears in reviews.) I’ve learned this the hard way by actually going to such a restaurant – the food was of course unremarkable, and at the end of our meal, the waiter unfailingly asked us for a social media review mentioning him by name. These kinds of restaurants are best avoided, in my experience.

Most of the time though, you won’t learn much from skimming the default set of reviews. Assuming you’re looking at a decent restaurant, you won’t learn very much – some people will love the food, others will think it just ok. Some people rave about the service, others will have issues with their waiter.

Look for the Unhappiest Reviewers

So next, I’ll sort the reviews from worst to best. The 1-star reviews often have people with very personal bones to pick about their experience, but if you see people consistently mention the same thing, that’s usually meaningful information. At this stage, there are more red flags you should be looking for.

Especially in Southeast Asia, it’s a near-guarantee that if the restaurant has thousands of reviews, you will find evidence of fake reviews once you start reading the 1-star reviews. Here’s how an actual restaurant I looked at in Vietnam is rated on Google Maps:

A screenshot of a Google Maps listing for a restaurant rated 4.7 with thousands of reviews

And here’s a review I found within a minute of scrolling through its 1-star reviews, talking about how you get a free drink in return for reviewing them:

A review of the same restaurant on Google Maps mentioning that you get a free drink for leaving a review

So once more, you can rule out these restaurants. A majority of the time that I’ve been excited about a well-reviewed restaurant and found the actual visit a thorough disappointment, it turned out to meet one or both of the red flags for fake/purchased reviews that I’ve discussed here.

Once you’ve learned some of their reviews have been bought, it’s almost certain that the majority of them are not reliable – so you might as well dismiss this restaurant altogether. It’s usually also a negative signal about the restaurant’s faith in their product when they have to resort to this method. That’s not to say all restaurants which do this are going to be terrible – there’s exactly one restaurant I’ve been to that buys reviews like this which I’ve still found decent. But every single other such restaurant I’ve tried has been somewhere between disappointing and terrible.

Most of the time, you hopefully aren’t finding any obvious red flags that the restaurant’s reviews are fake/paid-for. The 1-star reviews here are still helpful, especially if these reviewers are consistently let down by the same issue. They might all feel the value for money isn’t there, or maybe you’ll learn that the restaurant’s star dish is extremely polarizing. When multiple people mention that a particular dish is disappointing, or that the restaurant adds a mandatory gratuity, this is all useful to know if you do wind up going to this restaurant. And if any of these flags might be dealbreakers for you, you know to walk away.

I usually keep scrolling the worst-to-best reviews until I get to the 2- and 3-star reviews. A tip I’ve heard before is that 2-star reviews are often the most informative, because they clearly hated the restaurant but at the same time found something redeeming enough that they couldn’t give it just one star. This is generally true in my experience – the 2- and 3-star reviews are often where I find green flags that confirm a restaurant is worth trying. These reviews are often left by people with enough maturity and judgment to realize that the restaurant let them down on something they care about, but that someone else might not share their priorities: they are likelier than 1- and 5-star reviews to have informative nuance.

If the majority of 2- and 3-star reviews are talking about service issues and pretty much nothing else, I usually take this as a green flag. This often means the food is good, and the restaurant’s primary execution issues lie elsewhere. As long as there are an ample amount of 5- and 4-star reviews, the restaurant is likely perfectly capable of executing good service most of the time. And in situations where service is not a priority, that’s not even a problem: in the West, some (myself included) hold that the best Chinese restaurants will have a 3.5-star rating, because their service is just not going to be good by Western standards, and that’s fine. 

It’s rare, to be clear, that a restaurant’s 2- and 3-star reviews are going to all be about just one or two issues that you’re comfortable ignoring. Most of the time, they’ll raise a handful of issues and you’ll have to make the call for yourself whether you want to take a chance on this place. Maybe most of the reviews are happy with the food, but there are also quite a few 2- and 3-star reviews that find it overrated. You’re rarely going to find consensus; at this point, you’ve eliminated all the red flags you can, and you’re just going to have to make a personal judgment on what sounds worth trying.

Final Tiebreakers

If you haven’t seen any red flags yet, we’re in the home stretch. The final thing I do is sort reviews from newest to oldest. One issue that’s rampant now we’ve had the internet around for literal decades is that so much data is stale. A restaurant that served delightful food seven years ago might just be mediocre today; or it might have completely reinvented itself and be serving still-delicious food but with a completely different menu. No review platform I’ve yet seen has even tried to deal with this: reviews from a decade ago are given the same weight as a review left yesterday, even though most restaurants will have turned over their menu (or even owners) multiple times since then.

At this point, the main red flag that can show up is whether the restaurant still exists. The newest reviews can often warn of critically basic things like whether a restaurant has recently closed, temporarily or not (I’ve found surprisingly often that the only warning sign of this can be a recent 1-star review). On rare occasions, the newest reviews can also show if the restaurant’s overall high rating covers up a recent trend of weakening food quality. You’ll also sometimes glean intensely seasonal and contextual knowledge from recent reviews – perhaps a star dish is off the menu right now because the ingredients aren’t available, or the menu’s recently been changed.

If you’re looking to winnow down your shortlist of restaurants even further and have the time, it can be worth searching for reviewers with specific knowledge of the restaurant’s cuisine. When I scan reviews, I often look for someone mentioning they’ve lived in the country/region that the restaurant’s food is from, or a name that indicates the reviewer might be from the culture that the cuisine originates from. 

The reviewer’s name is not a useful signal in a highly diverse country like the US, but it’s been surprisingly helpful sometimes now that we live in Asia. In Malaysia, it’s a red flag for me if a well-reviewed Western restaurant doesn’t have any reviews from someone with a non-Asian name. The frank truth is that most Malaysians have not tasted authentic Spanish tapas, Mexican tacos, or American pizza, and so if that’s what you’re after, you will never be able to discern this from reading solely reviews left by locals.

A plate of Caesar salad topped with round rice cracker balls and seaweed flakes

A popular Malaysian rendition of Caesar salad, topped with seaweed and rice crackers – Malaysian reviewers on Google Maps love this, but if you’ve ever had a real Caesar salad, this will likely be an unpleasant surprise!

In Vietnam, the easiest way I found to confirm a Vietnamese restaurant was worth going to was to look for reviewers with Vietnamese names: I basically ignored all reviews left by non-Vietnamese and only skimmed the highest- and lowest-rated reviews left by Vietnamese people to form my opinion. The median foreigner won’t know what Vietnamese food is supposed to taste like, but the median Vietnamese person should. This tip works particularly well if you’re in a place where the locals are active on Google Maps – I found it less helpful in, say, Bali.

Even so, after applying all of this, I would say most of the restaurants we try are somewhere between disappointing and ok. But our success rate is increasing, and we’ve managed to find places we enjoy eating everywhere we’ve visited. The number of times we’ve had a truly disappointing meal and been surprised is certainly lower than ever. 

One thing I’ve found to be true both professionally and personally: taking a data-driven approach can’t guarantee you success, but it can certainly make you much more aware of how you might fail, and knowing is half the battle. Analyzing reviews along these lines has made me a lot more aware of which specific risks I’m taking when I try a new restaurant. While it’s boringly nerdy to boil restaurant selection down to a risk management exercise, it’s also really helped manage our expectations when trying an unfamiliar restaurant  – and I hope something here can help you too.

Reply

or to participate.