Successful media is entertaining. It has to be to get you to keep coming back day after day, many times a day.
Getting this right requires a mix of serendipity and high value content. Not enough high value content and I feel like I’m wasting my time, and I leave. Techcrunch had a post this weekend speculating that Twitter quitters are leaving because of a noisy feed. On the other hand, not enough serendipity and a service starts to feel stale, and I leave. I may as well be using an RSS reader. Useful, but not fun.
This is the core problem of discovery, balancing novelty with content that is interesting to me. It is a signal vs noise problem. I’m willing to see some things that are not interesting to me (noise) so long as there is enough interesting material to make it worth while (signal), and the “cost” of the noise is not too high.
Users always underestimate their desire for serendipity. They think they know what they want to see, but they don’t. Serendipity, in the form of breaking news, viral posts, “most popular” content, and randomness of all kinds always garners attention even though users will tell you that they don’t want to see it.
Most social media sites solve this problem with the feed. A “friend” or “follow” dynamic puts some initial filter onto the content of the feed. But as the site grows, the volume of content that can go into the feed quickly overwhelms what a user can consume. It has to put some filters on what makes it into the feed. This gets to a personalization problem. How does a site determine what may be interesting to me? And how coarse a filter does it apply? The tighter the filter (executed well), the more likely I am to get items that interest me, but the less room there is for serendipity.
By viewing this as a signal vs noise problem, you can see how different sites solve this problem differently. If you have a site where the cost of “noise” is high, you will want to keep a tight filter and make sure that most of the things in the feed are interesting. But if you have a site where the cost of “noise” is low, you can afford to keep a coarse filter and allow for more serendipity.
Longform video is the most expensive form of noise. When you watch a new TV show, it can take one or more full episodes before you know if you enjoy it. That can be an hour or two of time. So long form video tries to keep a tight filter and guide you towards shows it has high confidence you will like. This can take the form of suggesting different episodes of a show that you have watched a lot, or simply the most popular shows. But it never shows too many at a time.
Short form video is less expensive noise, but still noise. It can take a few minutes of watching a youtube video before you know if you like it or not; the average Youtube video is watched for four minutes. And once again, Youtube keeps a relatively tight filter, relying on subscribed channels and related videos for its recommendations
Long form text is the next most expensive noise. Examples would be most news, blogs, magazine and media sites. Read an article and you’re in at least 30 seconds before you know whether you’re enjoying it or not. Most such sites have 10s on links on their home page, the equivalent of their feed, and the editors spend a lot of time optimizing to determine which articles make it to the home page.
Short form text is pretty cheap noise. In the second or two that it takes to scan a tweet or Facebook status post to see if it is of interest to you, you’re already on to the next item in the feed. So it is no surprise that these are infinite scroll feeds within a single column.
Finally, if a picture is worth a thousand words, then it makes sense that picture based media is the cheapest noise of all. In a fraction of a second you can glance at a picture and know if it captures your interest. So it can have the coarsest filter of all. And that is why you see an endless scroll of photos in multiple columns in Pinterest, Wanelo, Whisper, Poshmark and so on.
The lower the “cost” of putting something potentially uninteresting into your “feed”, the more room there is for serendipity in content discovery. It allows the personalization filter to be coarser. That’s feature, not bug.