Flickr had it, Instgram has part of it, Facebook is close, 500px almost has it, but listen, there's a lot more to photos than simply "popularity." The most widely used algorithm, as far as anyone can tell, is based on a few simple properties:
- Quantity of views
- Quantity of actions taken
- Quality of person who viewed or acted (based on their own accumulated algorithm results)
- Time (recency)
This works great in the early stages of photography communities, where not one group has a huge lead over newcomers. When everyone has just about the same amount of followers, and almost all the same actual followers, any algorithm crunching the data above will surface some amazing photographs from everyone. As soon as you start seeing discrepencies that will easily leave newcomers in the dust, you start having a problem.
Photo discovery isn't about popularity. It's far from it. Photo discovery should be about surfacing photos based on how absolutely awesome it is. Composition, emotion, story, and reach. It's lazy and incorrect to just chalk up popularity on quantity and influence.
If Ansel Adams, who we should assume hypothetically isn't famous yet, joins Flickr today, he will not make it to the Explore page. Here's a few reasons: he's not in the group of Thai middle school kids who are incredibly influential now on Flickr, he's not taking pictures of hot chicks, and he doesn't post pictures of cute puppies... and most of all, he isn't posting unrealistic colorful HDR photos. The only way to have your photos discovered when first joining Flickr is to stand out as a thumbnail in a Flickr Group you've posted to. Best chance is he can post his photo in the B+W landscape only Flickr group. Actually I take that back, he should post a full color photo of a hot chick holding a puppy in the B+W landscape only Flickr group, tagging all the Thai middleschoolers in the photo.
On Instagram if you're a celebrity on Twitter, you're good to go. Otherwise you'll be buried and never make it to the popular page unless you have the qualities of a girl who does a ton of self portraits, or have a funky designer mutt that is named after household appliances (to be fair, kevin rose's labradoodle is pretty freakin cute). I know plenty of amazing photographers who never get on popular, yet I see iPhone note screenshots make it up there. Usually these notes are from girls or influential tech founders leaving tidbits of knowledge. yah, these are interesting, fine, but you're still not discovering the greatest photos. What Instagram does have right, is being able to see what photos people you're following are also liking and commenting on. I think exploring this route as something more primary than simply popular photos would really be beneficial. Probably not perfect and interesting as it is today, but what your friends have discovered and passing it on would be a step in the right direction.
500px has an editorial section that worked great early on to surface some otherwise unknown photographers into the limelight. Unfortunately, they just can't keep up with the scale of photographers now using the site. I've seen people who are influential on Flickr ask their Flickr communities to help them bump up their numbers in 500px. They need to fix this otherwise they will fall to the same demise of Flickr Interestingness woes.
Instead of going on and on about this, you can clearly see a pattern. Photography discovery has a long way to go, yet there hasn't been a lot of innovation on this front. You can't simply take simple stats to surface the best photos people will care about. There needs to be a level of heuristics involved. The problem is that there's no clear logic to figure this out just yet, and you can't just go 80% and expect it to work well, you need to hit at least 95% accuracy or higher, which is hard. People who work on natural language processing live in a similar world. If the NLP is only 80%, it can be impressive, but won't work enough for people to keep coming back. The last 20% is exponentially harder and seemingly even more time consuming.
Some good things to look out for: image recognition, meta information tied with current issues, crunching interests with photographs, and putting more emphasis on photo referrals from friends.
Image recognition can go far. It's similar to natural language processing, and I guess you can even call this image processing. If you can programatically come up with a way to know how awesome a photo is based on composition, quality, etc., then you can easily come up with a bunch of meta information to crunch into an algorithm based on interests and so on. This is a hard problem, Face.com is closing in on facial recognition, and Facebook has already implemented a easy way to suggest tagging faces in your photos. If you can take a few photo properties, like color depth, image size (if an image size is larger within the same dimentions as some other photo, you can probably accurately assume there's more detail and higher quality, which isn't always better but useful), see if it uses the Rule of Thirds and Golden Ratios, has boobs, has cute fluffy ears, you can really revolutionize photo discovery.
But until someone builds some robust image proccessing library, there's plenty of things to be done with suggestions from friends, newcomers who are influential in other networks, editorially picked photos (on a massive scale, like mechanical turk), and pulling out more meta information that's available. I haven't seen anyone do this just yet. Dedicate a team to photo discovery as well as the usual photo sharing/virality team. It is a huge factor in long term success and not just a passing feature you poke at once in awhile.
