A major dilemma for stock photo portals is finding the right size for their collections. Customers complain about too many irrelevant images. Thus, a small collection would seem to be better. But an irrelevant image to one buyer, may be relevant to another. Cut the collection size, and customers complain they can't find the right image.
It is interesting that Alamy.com, one of the fastest-growing portals in number of images licensed is also the fastest-growing in size, with more than 10.5 million images in the collection.
Many collections go through periods of adding images from a variety of sources to get diversity. Then they place arbitrary limits on the number of images suppliers can add to the collection and edit images out to reduce the collection's size.
Some tend to remove unlicensed images after they have been on a site for two or three years, on the theory that newer images will be better. That may be short-sighted, since the image is perfectly serviceable.
Everyone wants an automatic solution for getting the best images to the top, but for the most part, that hasn't worked. It may be time to recognize that no automatic system is going to make the kind of artistic judgments an editor can.
One solution, yet untried, involves a change in editing strategy. Accept everything except those with obvious technical faults. This makes all possible choices available to customers who visit the site, and frees up editors for more important work.
Focus the editor's time on organizing the search-return-order (SRO) of the highest demand subject areas, so the best images in the editor's judgment come up first.
Here's how I think it should work: Portals know which keywords are used in order of frequency. They know which images have sold, and the keywords used to find then. They know how many thumbnails the average customer looks at before doing a new search.
Pick the keyword most frequently used. Let's say it's "couple." On Getty Images, there are 139,786 couple images. Assume statistics show customers will go through 1,000 thumbnails, on average,Â before giving up. (Alamy says their average customer looks at almost 2,000 thumbnails). We want to get the best (based on an editor's judgment) of that 139,786 into the first 1,000 images shown. The editor needs to go through all 139,000 and code them, so the best will show up in a particular SRO. A hard look should be taken at images that have sold, no matter how old, because at least one customer found them useful.
Suppose the customer wants "couple on the beach" (13,909 on Getty) or "senior couple" (8,188 on Getty). These are more reasonable numbers but still more than anyone will look through. These groups still needs to have the first 1,000 ordered, but will be ordered differently from those with just "couple" as a keyword given the additional qualifiers. Some of the 139,000 that didn't show up in the broader search will be included in these groups because many of the images in the "couples" grouping no longer apply. If some keywords are used a lot, but few images licensed, it may mean that the best images are too far down in the pack.
It isn't necessary to do this for every search term. In some cases, the numbers of images found are so small that customers will probably not be inhibited from looking at everything available. "Repairman" on Getty had 1672 images, but Getty also breaks this category down into "Manual Worker" (530), "Mechanic" (1197) and "Maintenance Engineer" 156. There is no need to bother adjusting the search return order.
Going through large groups like "couple" will be time consuming initially, but updating on a regular basis shouldn't be that difficult. The editor reviews new images in the collection monthly and determines which ones should be added to his initial 1,000 and which should be replaced. Sales statistics will also be helpful here.
More and more portals tend to have many of the same images. Each can distinguish itself by the SRO it offers. Buyers in India, Poland or many other countries may have different ideas as to what qualifies as a useful image. Editors from each culture can create local SROs. Two different sites can have the same images, but customers will see a very different selection based on trends in a particular market.