Ratings Gone Wrong

Ratings Gone Wrong

Five-star rating systems are broken.

And for all the greatness of the internet, I can't believe we're still stuck with them.

Historically, the internet has been a gift for people with obscure tastes. I like Derek Thompson's framing: the internet is like Tokyo. The Japanese city is famous for all kinds of strange shops, which make economic sense in a city with 40 million people, but not in a small city like Des Moines, Iowa. That's why, on the internet, you can be niche at scale.

From dating to Airbnb apartments to Pez Dispensers on eBay, the internet is an incredible matching machine. A newsletter like Monday Musings works better as a global email newsletter than a local neighborhood magazine. If the internet wasn't so unbelievably massive, I wouldn't write it.

For reasons I've discussed before, mass media catered to the masses. That's why the commerce companies who advertise on mass media focus on the averages, not the extremes. The internet is different. It's a gift to people with obscure tastes and strange obsessions.

Unfortunately, five-star ratings cater to a world of averages. By de-risking our experiences, they eliminate the potential for the kinds of strange and spectacular experiences that are so memorable. Worse, rating systems don't transfer across platforms. Intuitively, we know that a 4.6 on Uber is much worse than a 4.6 on Yelp. But unfortunately there's no way to know how ratings are valued when you join a new platform.

Should we trust the 4.1 on TaskRabbit? How about the 4.8 on Handy?

To illustrate my point, I'll focus on food, books, and travel.

Food: Most of my favorite restaurants are quirky and weird. They're the kinds of restaurants that would turn off some people. They look nothing like the recent wave of fast-casual chains, which are optimized for modern rating systems. Instead of enabling memorable and one-of-a-kind experiences, 5-star rating systems encourage restaurants to play it safe and eliminate all the potential for downside risk. No matter where you go, following a 5-star rating system will lead you to generic coffee shops with the same wooden chairs and minimalist design. I call this The Algorithmic Trap. In an ideal world, restaurant ratings would change throughout the day because a 5-star restaurant that's packed for dinner is usually a worse experience than the same one that's empty for lunch.

Books: Many of my favorite books are loved by a small group but disliked by most people. They're usually weird and obscure. To put it in terms of ratings, they have tons of 1-star reviews and a handful of passionate 5-star ones. Unfortunately, the Amazon five-star ratings system hurts these books and makes them impossible to find. That's why the 5-star rating system on Amazon is so unhelpful. To Amazon’s credit, they’ve made massive strides on social recommendations. The “people who bought this also bought” section has improved tremendously, but still doesn’t yield as many surprises as it should. Until something changes, I'll keep relying on friends and footnotes for book recommendations.

Travel: Scroll down TripAdvisor and you'll see a handful of mainstream experiences. Every city has the same mix of walking tours, bar crawls, and open-top bus rides. Borrrrrrrrrrrr....ing. However, there is one exception. To their credit, Airbnb incentivizes hosts to rent out quirky homes by featuring them on the home page. Instead of focusing on ordinary averages, Airbnb highlights the crazy extremes. That's why I love the platform so much. More of that, please.

Screenshot 2019-09-08 22.20.14.png

Whenever I analyze a system, I look for the incentives that drive it. As Charlie Munger once said: “Never, ever, think about something else when you should be thinking about the power of incentives."

The world needs rating systems that lead us towards eclectic experiences. Unfortunately, the incentives of 5-star rating systems inspire bland experiences. Maybe, rather than a universal rating for a given experience, the software could adjust to your specific preferences. After all, one person’s 5-star is another one’s 2-star.

Ratings systems could also give more weight to friends or trusted influencers. Or, perhaps the software could highlight low or high-variance experiences that reflect the user’s risk appetite. If I’m in a big group, I want certainty. But if I’m traveling alone, I’m willing to put up with some flops in exchange for epic adventures.

Until the incentives change, we'll be stuck with experiences that are too safe to be exciting and too generic to be memorable.


Note: If you’d like to receive future posts by email, subscribe to my Monday Musings newsletter.