Organizing the flow of images and extracting value. Finding, in the millions of images being shared daily, those that have a potential to be sold. How do we figure out this riddle ?

In a time where everyone is talking about big data and how to make sense and profit from  huge amounts of information, the answer should be easy. But it is not.

Professional photo licensors, agencies or individuals, have been working on this question for decades, albeit with lesser volumes. Using a mix of human psychology, sociology, aesthetics, trends and past sales, along with gut feelings, they have somewhat successfully offered, and continue to do so, an acceptable curated collection of images that has somewhat satisfied the demand. Somewhat because without offers, it is hard to know if there is a demand and it is quite possible that they have unselected images that would have otherwise been bestsellers.

Phase two has been what we experience now with the mega libraries offering tens of millions of images. The approach here is to lower both the bar of curation and pricing, while leaving the decision to the buyers. In this model, the decision process of what is being offered, still left to the care of human beings, is in obtaining critical mass – that is enough photographs to satisfy all potential needs. Much lower cost of inventory makes this acceptable but leaves millions of offered but  unsold assets.

But even that solution doesn’t come close to solving the current issue. It can only work with a pool of voluntary submissions and leaves the vast majority untouched. Those methods, however seemingly efficient they might appear, cannot be transposed to working on all images uploaded. Not even google incredible indexing power can keep up with the pace of todays overall photo  uploads. Just think then if every image had to be reviewed by individual editors, as it is still currently done. Impossible.

Light PaintingCurrent thinking is to put in algorithm what the human editors do and pass the bucket to fast processing machines. While tempting, it is immensely difficult. Why? Because are understanding of aesthetics is highly volatile and open to large variations.In fact, it has to in order to keep current. We do not like the same images as we did in  the 70’s and even 3 years old image look, well, so last year.

Putting human curation ability in a box might actually be a bad idea. No numbers exist to confirm this but it might just be that it is  very inefficient. It might just work like the book publishing world where one successful book compensates for the hundred that hardly sale. The 80/20 principal. 20% of images in a collection are responsible for 80% of the sales. Why would we want to replicate this model and make it even faster ?

The beginning of a solution is elsewhere. One, that has shown little success up to know but shows promises is taking the problem upside down. Instead of presenting an offer and hope it matches a demand , take a demand and find an offer. Companies like ImageBrief are currently applying this model. It doesn’t rely on machine curation, but rather two sets of eyes, the buyers and the photographers, to succeed. Still very inefficient as thousands of images offered remain unsold. But the good news is that those unsold images do not to be stored endlessly. And the buyer doesn’t feel he’s in a bad yard sale, having to scavenge through thousands of useless items.

Another is to anticipate demand and find matching offers. Crowdmedia does that by locating Twitter image uploads that match hot news items, contacting the owner and offering the photos for licensing automatically. The only curation here is the nature of the news event and the willingness of the photographers to put their images on the market. Brilliant. But limited to news. But it is a step forward into anticipating the markets demand and then, only then, looking for matching content. Imagine if we could expand this model for all demand and being able to search every single image uploaded daily ? Anticipating an upcoming need for a certain type of images and then scouting Instagram, Flickr, Facbook, Twitter , Tumblr and so on to find the best matching images, contact the photographers and offer them to license the images for them, and upon approval, put those images in front of potential buyers. That would be an incredible photo licensing platform. Sounds impossible ? Well think about this. Currently, Wall Street companies are spending hundreds of millions of dollars to create algorithms that can  anticipate  stock market fluctuation. They scout information freely available on the internet and process them in order to detect trends, news, variations before they happen so they can buy or sale stocks automatically . All this billions of time a second and all day long. The same computing power could be applied to the photography market. When it becomes cheaper, obviously.

We are not that far. It would be rather simple today to create a strip down version of this with celebrity photography for example. By mashing gossip news ( and there is plenty available), extract who is in and will be in demand and then scout the internet for pictures of that celebrity to put on the market. Facial recognition software are sufficiently advance to do this. From there, expand to travel photography.  Aggregate information from travel sites per season. And so on.

The current big mistake is to believe that the value is in the offer. That if you build a wide enough collection of images, you will hit the winning numbers, or close, enough times that it will pay your bills. It is not. It is in the demand. More precisely in matching exactly the demand . Too many companies start by creating a library of images and then go out and seek a matching market of demand. It is wrong and prone to failures. The answer is into cornering a demand and then creating an offer.

Author: pmelcher


Leave a Reply

This site uses Akismet to reduce spam. Learn how your comment data is processed.