Have you noticed that when a customer searches for “man” on some major agency sites a different number of images will be returned than if they had searched for “male.” An agent called this to my attention after hearing complaints from a number of customers about this problem when searching on GettyImages.com.
Most agencies use “word trees” as part of their search algorithm to make sure that regardless of the word or word strings customers use they will find all the relevant images in the collection. If a customer uses a singular noun the search engine will also deliver all the images with the plural of that word in its keyword list, and vice versa. When the customers search for “man” in many cases they will also get all the images with the keyword “male.”
I decided to run a small test to see how well some of top agencies are doing.
|
Getty Images |
Alamy |
Corbis |
Shutterstock |
iStock |
man |
1,835,436 |
4,607,656 |
3,014,569 |
6,268,439 |
3,139,502 |
men |
1,808,665 |
2,056,477 |
3,015,033 |
6,269,350 |
3,139,502 |
male |
2,337,527 |
3,332,225 |
2,298,417 |
4,169,832 |
1,943,876 |
males |
2,249,052 |
405,029 |
2,297,937 |
4,169,806 |
1,943,030 |
|
|
|
|
|
|
child |
816,506 |
1,783,652 |
1,122,009 |
2,809,540 |
1,597,269 |
children |
617,544 |
1,279,594 |
1,122,013 |
2,810,151 |
1,668,848 |
kid |
803,281 |
784,861 |
1,295,917 |
1,998,687 |
1,599,459 |
kids |
571,784 |
556,505 |
701,229 |
1,997,938 |
1,599,459 |
|
|
|
|
|
|
happy child |
304,329 |
533,193 |
187,458 |
1,964,730 |
797,994 |
happy children |
333,750 |
271,517 |
187,462 |
1,673,725 |
826,158 |
happy kid |
303,381 |
329,579 |
192,609 |
1,964,781 |
798,396 |
happy kids |
302,937 |
145,949 |
182,348 |
1,673,466 |
798,622 |
In most cases the count for singular and plurals seem to be about the same with Alamy having some big discrepancies and Getty Images also having some problems. Corbis is good on most words, but when it comes to “kid” and “kids” there is a big discrepancy.
There is a bigger problem with words that often have the same meaning, but also can have meanings that are slightly different. Take “male” for example. It can mean the same as human “man,” but there are also male animals and other species where the word male might be appropriate. The word “kid” is sometimes used interchangeably with the word “child,” but it also can mean a baby goat.
From the creator’s point of view, it is important to be sure that all -- not just some -- of the relevant keywords are attached to your images. There is no way to anticipate what words a customer will use when searching for a particular subject. If your agency doesn’t have an effective word tree as part of its algorithm, then you need to make up for it by adding all the appropriate singulars, plurals and other keywords.
This is further complicated by the fact that many agencies encourage contributors not to use a lot of keyword because that can slow down the overall search. In addition, sometimes if a word is added that relates to a small or minor aspect of an image it may result in delivering a lot of inappropriate images for customers to review.
Most agencies seem to think that if they deliver tens of thousands, or even millions, of images on a particular search, no customer is going to have time to look at them all anyway. Thus, it doesn’t make much difference if some of the best images on the subject are not included in the search return. For the agency it’s only important that the customer buy something. This may not be in the best interest of most image creators.
Based on the chart above, it’s hard to explain the differences between the iStock and Getty Images. Based on this very limited sample iStock is certainly the best with Getty being fourth out of five as a result of some big discrepancies. But, it is my understanding that the technology departments of both iStock and Getty Images are now part of one central organization in Seattle. And iStock has about twice as many images as Getty has in the creative section of its site. If iStock has worked out an efficient word tree system, why isn’t Getty using it?
It would certainly help everyone if creators – who are doing the keywording – were given more information about what words customers are actually using to search for images, how frequently they use any given word and which words, or combination of words actually result in sales. The data is there, although it may not be captured by every agency. But the agencies have decided that this information is proprietary. They don’t want to disclose it to their competitors. As a result everyone is blind and making decisions based on guesswork and totally inadequate information.
This leads to more, unhappy customers. There are some early indications that some of the higher end customers are turning away from stock photos because it is becoming harder to find quality images that everyone else isn’t using. These buyers are going back to hiring photographer to shoot what they want because finding the right image on a stock photo site just takes too much time. If this is the beginning of a trend, it is something to think about.