Search Results Lottery

10

March 10, 2006

Search result position is critical. The higher your image is in the results delivered after a customer has completed a search the better the chance it has of being seen and selected for use. Thus, it is important to understand how search results positions are assigned, and the factors that impact a change in that position.

Most search engines operate on the principal of the last-image-uploaded-appears-first. Each image is assigned a date of upload. The software searches the date field to determine which image of all those found to show first. Thus, the longer an image stays on a site the further down in the return pile it gets, as the date gets older, the less chance that it will ever be seen. This is particularly true if the subject matter is very generic and there are many such images on the site.

It is also important to note that the age of an image does not necessarily mean the date it was taken. A picture might have been taken in 1990, but it wasn't loaded on a particular site until September 2005. In this case any image that was loaded earlier will appear below the 1990 image even if the new image was actually shot in June 2005. Another factor that may be used to influence search result position is whether a high resolution version of the image has ever been downloaded or whether a customer has actually purchased the image. This could bring some of the older, but more popular images to a higher position in the search results. As far as I can tell no one is currently using this factor in returning search results. I understand that Getty Images is considering using sales history as a factor when determining search order, but so far has not done so.

Another techniques that I understand Getty Images sometimes uses is "resurfacing" certain images. This is a process where the date the image was originally entered into the database is changed to a more current date making it appear that the image was loaded much later than it actually was. This is particularly useful for timeless images and concept pictures where there is no way to identify the date they were shot.

With this preamble let's look at the Getty web site to see just how frequently images from certain brands appear. Recently a number of Getty photographers and Image Partners have complained of declining royalties and attributed this fall off to changes in the position in the search return order. I want to examine this issue.

For my initial analysis I searched for all images, both RM and RF, in the database using the term "vertical" on the theory that all brands have some verticals and thus this search would give the best understanding of the order and priority given each brand and the relative percentage of the total images for the brand.

There were a total of 553,841 Vertical images. My chart shows whether they were RM or RF and the number of images each brand had in the first 90 images delivered from the search. It also shows in the fourth column at what point in the sequence the first image of the brand or image partner appeared, and finally the total number of each brand's images in the first 270 or 540 shown. In theory 540 images are a lot to search through in order to find something you can use, but in fact many customers may be willing to search through a lot more than that.

It is also worth comparing these numbers with those in Story 578 published in September 2003 and Story 681 published in November 2004. Story 681 is particularly interesting because I also searched on "Vertical" in that story and the chart provides us with a basis for comparison with 15 months earlier figures for the first 90 and 270 images.

In comparing November 2004 with the current numbers we see that there has been a significant drop in the numbers of virtually every brand's images that are now being seen. This is true for the Getty owned brands as well as the Image Partners. For example the number of images delivered in the first 270 and the percent decline for a few of the brands are as follows:

The only two brands that stayed even or improved are Stockbyte and Photographer's Choice. In Stockbyte's case they had 8 images in the first 270 in 2004 and they still have 8 images, but now they are is split among 4 different Stockbyte brands. In the case of Photographer's Choice (PC) there were 14 images in 2004 and now if we count both Photographer's Choice RM and RF there are 15 images.

Recently, many of the PC photographers have complained that their royalties have been dropping rapidly. Some have speculated that this has occurred because PC images have been given less favorable positions in the search return order as a result of the addition of new brands. It is significant that actual figures indicate that PC images have been allowed to maintain a MORE favorable position than the images of any other brand. This is particularly true when we consider the total number of images PC represents.

PC has 32,698 images on the site out of a total of 1,306,571 for all brands. Stone on the other hands has 109,171 images, 3.33 times what PC has, but Stone gets only 20 positions in the first 270 and PC-RM gets 12. If both were getting representation proportional to the number of images they have on the site Stone should be getting 40 while PC is getting 12. PC photographers have also complained about Photonica images eating into their sales. Photonica has 103,683 images on the site and again PC has much more favorable representation. On proportionality all the brands have a claim against Stone+ which only has 15,539 images at present.

Clearly brands do not get proportional representation relative to the total images each has on the site. But, if we look at the brands owned by Getty and the total number of images they represent compared to the number of images from Image Partners the share of search order positions they represent the overall share seems fair. There are 18 Getty owned brands on the site and combined they have 911,694 images. The 56 Image Partners have a combined total of 394,877 or about 30% of the total. When comparing the number of positions in the first 270 held by Image Partners with Getty's positions we find that Image Partners hold 29% of the positions. In the first 540 they have almost 34% so the Image Partners seem to be getting slightly more position the deeper in the search order we go. (Note: The sequence order for all brands seems to be repeated at some point between 200 and 240 images, thus, the fact that I did not break my search analysis at these points may result in a slight inaccuracy on these percentages.)

Virtually all of the other Image Partner brands had a 50% drop from 2 images 15 months ago to 1 image today. This drop is directly related to the number of Image Partners that have been added. As each new brand is added a position must be taken away from an existing brand in order to give everyone some representation. Given the rapidity with which new brands are bring added most of the long standing brands will continue to lose positions. All indications are that additional brands will continue to be added.

The Image Partners with a slight advantage are the 15 that have at least 1 image in the first 90. However, this advantage is very slight because most don't fare that much better than the others when it gets to 270 or 540 images. Fifty of the image partners have only 1 image in the first 270 and 49 have 2 images in the first 540.

This illustration demonstrates how little chance anyone has of getting an images seen when the subject matter is one where there is very high demand and every brand has images of this type to offer.

High Demand or Niche Images

One of the problems for producers is finding subject areas to shoot where there isn't already a huge over supply of imagery available. To illustrate the point I started by doing a search for "woman and laptop", a high demand subject area with lots of images available. This search returned 10,700 images.

Here we can begin to see that some brands do significantly better than their average position in the "vertical" search. Part of the reason for this is that 20 brands have no images in this category, so all those who do move up to a higher level. When the software encounters a brand with no images it skips to the next brand.

Note that National Geographic, which is very high in the search order when they have images, and has the potential to get 17 places in the first 270 and 34 in the first 540, gets only 2 images with this search. That's because they only have two images with a woman and a laptop. Those two images are shown in Geographic's first two positions, but when it comes to Geographic's third position it is skipped over and the search moves to the next brand in the search order.

It is important to note here that even with some very popular subjects some brands get a fair number places in the return results, but the seven major Getty brands - Stone+, Stone, Photonica, Taxi, The Image Bank, Iconica and Photographer's choice get the lions share of the positions. The Image Partners get to show very few of their images of high demand subjects.

If a Partner has three image positions in 540 that means when certain keywords are entered only the last three images that partner has loaded using those keywords are likely to be seen. Great images of the same type of subject matter that were loaded earlier have little chance of being seen.