Why does YouTube bother with ratings?

I’ve worked on a few online video sites in my career, and on every one of those projects we confronted the question of whether to have ratings. In this web 2.0 world, the answer seems obvious on the surface: Of course, you have to let your users rate the videos!

Kyte, for example, added ratings to their platform right before I joined the company.

But let’s take a step back and think about why we rate things on teh interwebs. On sites like Amazon and Netflix, we rate things to help their robots give us better recommendations. Our ratings are one input into a mechanism called colaborative filtering. Associating our ratings with other data – like product details (media type, author/director/artist, genre, price, etc.) and broader purchase metrics (other people who liked this also liked…), those sites try to predict what else we might like to buy or watch.

On sites like Yelp, we rate things for two reasons: The first is that we want to reward or punish… businesses in Yelp’s case in order to improve the overall ecosystem of businesses out there, and the second is that we want to participate in a system that helps ensure our own satisfaction as customers of said businesses (i.e. we can choose to patronize only those businesses that have high ratings).

But what about YouTube? I suspect people rate videos on YouTube for reasons similar to Yelp. That is, we want to congratulate people who produce good stuff and punish people who produce crap, and we want to guide everyone else (and perhaps YouTube’s editors) toward the stuff we liked. When you look at the numbers, however, people don’t appear to think rating videos is very important.

Looking at 20 videos from my personal list of favorites, I found that on average just 3 out of every thousand viewers bothered to submit a rating. And there was not a lot of variance. Most videos were very close to the average, so in this case 20 videos is a sufficient sample for a cursory analysis.

It would be interesting to know what percentage of the view count belongs to registered users of YouTube, since you have to be signed in to rate videos. I suppose it’s possible that just 0.3% of visitors to YouTube are registered and signed in, and that would fully account for the disparity between views and ratings.

But there were a couple of outliers that suggest otherwise. Of the 20 videos I looked at, two had elicited twice the average number of ratings. One is a virtuoso piece of editing work (not to mention research), and it represents a point of view that many people passionately share, so the higher number of ratings makes sense. The second video is by Don Hertzfeld. It’s not one of his best (and oddly it’s the second of two parts), but he’s a hero to those who know his work, so his fans have an interest in promoting him.

Twice the average number of ratings is still only 0.6% of view count, however, and that’s dismally low. The bottom line is people aren’t interested in rating the videos they watch.

Which brings me to a final observation. You almost never see videos on YouTube that have a high view count but a low cumulative rating. High views almost always means 4-5 stars. The only exceptions I could find were videos that elicit a highly polarized response – with political content for example. These end up with a 2-3-star average which doesn’t really reflect the true spectrum of mostly 1 and 5 star ratings.

So no matter how you evaluate it, ratings on YouTube are not a useful way for anyone to discover interesting content. The useful metrics are views and most-favorited. View count tells you what the masses are watching, and most-favorited tells you what people really found compelling (enough to want to watch again).

So, given this, if you were to launch a YouTube-like site, would you have ratings or not?