Frequency and User Ratings

Many social networking sites provide a user rating function that allows users to rate content. In many cases those ratings influence content ranking. A good example is YouTube, which uses a five-star ranking system and ranks content using some combination of ratings, total views, and total ratings (I do not know what the actual criteria is).

In these ratings systems for any content item the user is presented with two pieces of information: average rating and number of votes. Average rating tells the user how viewers who have taken the time to rate the video feel on average about the content. Number of votes tells the user how many people have taken the time to vote, which may influence how they value that average rating.

What is missing, and what I would love to see, is vote frequency data. I want to know how the average was reached. This would tell the user a lot more about how the community feels about the content in question.

I’m interested in this data in two categories. The first is votes as a percentage of views. Did the content in question prompt people to rate it more frequently or less frequently than the average for the site in question? A higher than average vote frequency would tell me that, good or bad, this content provokes a response. A lower than average vote frequency would tell me that the information was viewed more passively.

The second category I am interested in seeing is the frequency of votes for each of the five stars. You can reach a 3 star rating in three basic patterns:

• The majority of voters gave the content a 3 rating, while a small percentage voted higher or lower. This tells me the content in question prompts a lukewarm response. It prompted people to vote, but they didn’t feel strongly about it. This would look like a curved line in the shape of a frown on a graph.
• The majority of voters gave the content a 1 or 5 rating, while a small percentage rated the content 2, 3, or 4. This tells me that the content is polarizing; that it prompts a strong response from different groups within the larger site community. This would look like a curved line in the shape of a smile on a graph.
• Voting was equally distributed across all the ratings options. This would look like a straight line on a graph. I’m not sure quite what this would tell me, but it would make me think.

A high or low star rating would manifest as a steep curve, running either down left to right for content that the community did not like, or up left to right for content the community responded positively to.

In a user interface showing the vote totals would be awkward. But a graphic representation, something simple like the five-star visual, would tell you which of the five vote patterns led to the final rating. This could be a simple line curve or something that looked like an audio EQ graphic.

In my focus group of three colleagues is any indication, however, I am unlikely to see this type of information any time soon (think a sharp curve running from the top left to the bottom right). Consensus is that users won’t care; it is more information than they care about.