How TheKnow works
The technology behind personalized restaurant recommendations—and why it's better than crowd ratings.
The problem with crowd reviews
Traditional review platforms like Yelp and Google Maps aggregate ratings from thousands of anonymous strangers. A restaurant's score is an average—a single number meant to represent everyone.
But taste is deeply personal. The tourist who wanted fast service and the food critic seeking an authentic experience rated the same restaurant, and their opinions got blended together. The resulting 3.8 stars tells you almost nothing about whether you will enjoy it.
Worse, crowd ratings are easily gamed. Fake reviews, rating campaigns, and recency bias all distort the average. That new restaurant with 4.9 stars? Probably just hasn't collected enough diverse opinions yet.
A 4-star restaurant on Yelp might be a 2 for you—or a 5. The crowd average can't tell you which.
Our approach: Taste matching
TheKnow takes a fundamentally different approach. Instead of showing you what everyone thinks, we find people who think like you—and show you what they love.
Think of it like Spotify for restaurants. Spotify doesn't recommend songs based on global popularity. It analyzes what you listen to, finds other users with similar patterns, and surfaces what they enjoy. TheKnow does the same for dining.
When you rate restaurants, we compare your ratings to every other user. Statistically, we're looking for correlation: when you rate something highly, do they? When you rate something poorly, do they? Users whose ratings consistently move with yours are your taste matches.
The algorithm explained
Four key concepts power TheKnow's taste matching engine.
Pearson correlation
TheKnow uses Pearson correlation, a statistical measure of how two variables move together. A correlation of 0.7 (displayed as 70%) means strong alignment; negative numbers mean opposite taste.
Minimum shared ratings
TheKnow requires at least 5 shared ratings before calculating a taste match. With fewer data points, random noise could make unrelated users appear correlated.
Time decay
Recent ratings matter more than old ones. A restaurant you loved last month is more relevant than one from three years ago. Time decay keeps recommendations current.
Confidence weighting
A 75% match based on 20 shared ratings is more trustworthy than one based on 6. For You scores reflect both strength and reliability.
TheKnow vs the alternatives
See how taste-based recommendations compare to traditional crowd ratings.
| Feature | TheKnow | Yelp | Google Maps |
|---|---|---|---|
| Rating source | Your taste matches | Crowd average | Crowd average |
| Personalization | Per-user matching | None | Location-based only |
| Algorithm | Pearson correlation | Popularity/recency | Relevance/proximity |
| Accounts for your taste | No | No | |
| Gaming resistant | High | Low | Low |
| Best for | Finding restaurants you'll love | Checking general reputation | Finding nearby options |
Privacy and transparency
Your ratings power the algorithm, but your data stays yours. TheKnow never sells personal information to advertisers or third parties. Your ratings are used only to calculate matches and generate recommendations.
Taste matches see your ratings for restaurants you've both rated—that overlap is what creates the match—but you control what else is visible on your profile.
We believe in transparency. The algorithm isn't a black box: this page explains exactly how it works. If you have questions about your data, our privacy policy has the details.