Thoughts On Ways To Improve Search

As image databases become larger and larger it becomes more and more difficult for customers to quickly find the “right image” for their projects. Increasingly, customers are frustrated by this problem. They want more choice, but they don’t want it to take them longer to find an image they can use.

Often, in order to find something useful they are forced to go through too many similar images of situations that are of no interest to them at all. As search becomes more time consuming, customers feel they may be missing some very good images that are buried in the search-return order. This can be very frustrating.

Search would be improved dramatically if it were possible to search through a collection of previously licensed images separate from those that are new or have never been licensed.



If they could review only licensed images, as separate from the entire collection, the customer would know that at least one of their peers has found the image useful. This could be an asset not a liability. They would know that another buyer has found this image to be one of the best in the collection and that is it not just the photographer or stock agency trying to push them to buy something.

Given the huge numbers of images available in collections these days, the first decision facing most customers is some reasonable way to limit their choices. Most are not looking for images that have never been used by anyone else.



If the customer wants newer images, or images that have never been used, then she could easily toggle to the “Unlicensed Images” collection (assuming such a collection were available). In all likelihood will be 5 to 20 times larger than the collection of images that have been used.

Making it possible for customers to easily move back and forth between collections of images that have been used and those that haven’t could dramatically reduce the time it takes many to find a usable image.

In the collection of images used, images could be delivered based on the number of times used, gross revenue generated by the uses, or a combination of both. In such an event the most popular images would be shown first and the customers would understand that is the case.



In some of the current algorithms “Popularity” is based on clicks, likes or download, but since the seller’s purpose is to license images the best way to gage popularity is when someone actually pays something to use the image. In addition, on most existing sites, when a customer searches on popularity they are shown every image the collection has, even though most of the images have never been popular with anyone. The unsold ones are just lower in the search return order, but no one doing searches has any idea where the unlicensed ones start.

All this is simple enough for images that have been licensed, but what do we do about all those great, new images that have never been licensed? This gets us to the “Unlicensed” category.

Unlicensed

First, with this group of images we need to find a way to enable the customer to quickly review as broad a cross section of images as possible within a particular “keyword” category.

To do this the algorithm would first search all the images delivered in a search return for similars (99% similar) from the same shoot. Initially, only one of these images would be shown regardless of whether there were 5, 20 or 500 of the same situation. This would provide the customer with the widest possible variety of situations to choose from.

If the customer finds an image that is more or less what she is looking for, she can then click on a link that will take her to everything the photographer has produced that contain the particular keywords. At that point she would be able to review all the similars the photographer has produced plus any other images produced by that photographer containing the same keywords.

Another option when the customer finds an image she likes would be to click on a link to the agency that represents the image. That would allow her to review not only the images produced by this particular photographer, but by any of the other photographers represented by that agency.

Still another way for customers to search Unlicensed images would be to allow them to look at the newest images uploaded, based on the date of upload. However, the customer should also be able to specify the newest in the last 7, 30 or 90 days and possibly the last year.

Currently, with most search engines the “Newest” search returns every image in the collection in order of the date uploaded. If there 10,000 images in the search return the customer has no way of knowing if 1,000, 100 or 10 were uploaded in the last month. Some of the “newest” at the bottom of the pile were uploaded 5 of 10 years ago.

It the customer just wants to search the entire Unlicensed collection, regardless of the date of upload, then the search return could be organized based on clicks or being placed in lightboxes. But, it would be helpful if there were an explanation somewhere as to how these unlicensed images are organized.

Since we’re talking about unlicensed images a “fair search” might be to show one image from every contributing agency before showing the second from each agency and then the third and so on.

The goal in all of this would be to give each customer the best possible understanding of what is available on the site for any particular keyword set after the customer has reviewed just a few hundred images, no matter how many total images can be found using the keywords.

It would be important to make the workings of the search algorithm as transparent as possible to both customers and creators. The customers would then have an understanding of what they needed to do in order to find the images they need as quickly as possible. Creators would have a better idea of what customers are actually purchasing and what they need to be producing to fulfill customers future needs.