Corbis Updates Search Engine

If you haven’t visited the Corbis web site for a while it may be time to take another look. In 2010 Corbis decided that it was time to update the search technology they had been using since 1995.

The project involved a series of architectural changes with two primary aims – improving performance (speed of delivery) and relevancy of the search results delivered. They examined a number of search engines and settled on Lucene/Solr which is an open source enterprise search platform from the Apache Lucene project. According to Kathy Golden, Corbis Senior Director of Software Development, their first requirement was that the engine be “an off the shelf, proven software that is used by lots of top companies.” The software also needed to allow them to get in under the engine and make customizations to deal with the search situations that are specific to Corbis.

With the search architectural changes they were able to get a 2.5 times or 250% improvement in the speed of delivery of search results. The next challenge was to deal with how the old engine parsed keywords. It looked at phrases like “tiger woods” and provided pictures of tigers and woods as well as the person Tiger Woods.



Golden said, “We examined the trends of customers search behaviors on search engines and found that today they tend to type sentences rather than just a few words. They want the engine to be smart enough to recognize when phrases or side-by-side words have a different meaning than each word individually.”  Fortunately, this didn’t require a change in the way they keyword. They weren’t changing the words they were simply changing the relevancy of what the engine was giving back. Customers have been enthusiastic about the changes and in the first month after the initial launch Corbis had an 182% increase in the number of searches on the site.
 
Next, Corbis did extensive interviews of image buyers. They discovered that buyers prefer a system on the left hand side of the search screen -- similar to those being used by Getty Images, Alamy and many of the microstock site -- that makes it easy for them to refine their general searches. This is usually referred to as “guided navigation.”



To make searching faster and easier for the customer Corbis included an “Auto Complete” feature that provides suggested words and phrases related to what the customer is typing. Clicking on one of these suggestions can save the customer typing time and speeds the search process.

The two major subdivisions of imagery are Creative and Editorial. In its Creative category Corbis offers the standard RM and RF options, but given its extensive and wide ranging editorial coverage it breaks down Editorial into: Documentary, News, Sports, Entertainment, News Archive, Historical, Fine Art and Outline, their exclusive celebrity portraiture brand. With one click customers can choose to search any of these categories.

Customers can also narrow their searches in a number of other ways including: Number of people, Gender, Age, Ethnicity, Type (illustration or photography), Orientation, Point of View, Date Image Added to the collection, Releases and whether or not an RF image is available on a Virtual CD.



Corbis offers 10 different age breakdowns and 7 general Ethnicity breakdowns. Both Alamy and Getty Images offer more extensive breakdowns of these categories. There is some question in my mind as to whether customers really want to know whether a child is 6 or 8, or if like Corbis does, the “Child” sort is for ages 3 through 9. Then it is left it to the customer to determine if the child looks the right age. Alamy lets customers choose from almost 80 different ethnic groups, but if the customer is really interested in that degree of specificity she can easily enter the word for the specific ethnic group as a search term.

One of the trade-offs in guided navigation is finding the right balance between too few choices and providing so many choices that it becomes confusing. Corbis seems to have struck a nice balance in this regard. Another factor to be considered is that the more options there are the more likely the keywording of some images will be inaccurate. This is particularly true when the image creator, or the Media Partner, is asked to supply the keywords.

One feature that both Getty and Alamy offer, which Corbis does not, is a count of the number of images in any sub-category. For example if you do a search for People on Alamy you are told exactly how many images of 4-5 year olds there are before you actually click to narrow the search. On Corbis you must actually complete the search before the number of images is revealed. Corbis could improve their offering by adding the image count by each sub-category.

A unique feature that only Corbis offers is whether an RF image is available on a Virtual CD. Customers have indicated that they like to know if the images they are choosing are available by purchasing a CD.

Corbis still offers an “Advanced Search” feature on its site, but says that less than 2% of customers use it. Most of those who use advanced search are trying to search specific collections and Corbis is exploring ways to make that process simpler.

Corbis Collections

Corbis represents at least 114 brands, both its own and Media Partners, and it has divided them into 4 price categories: Premium, Standard, Value and Super Value. Within each of these categories there are some variations in pricing depending on the brand. At one point all images represented by Corbis exclusively were priced at the same level depending on how the image was to be used. That is no longer the case.  

About 33 of the brands on the site are Corbis created brands. The rest are from Media Partners such as Blend Images, Image Source, National Geographic, etc. In some cases the Image Partner brands have images at multiple price points and some sell both RM and RF.

In some cases Corbis created brands are made up of images they have received from Media Partners. For example the Ocean brand is made up largely of images from PhotoDisc. (Corbis is one of the distributors of images belonging to Getty. It is unclear if Getty licenses rights to any Corbis images). It is unclear whether all the images in Ocean come from PhotoDisc or whether images from other Partners are included as well. One of the theories behind the separate collections is that images with a common style have been grouped together in one collection. For example, in theory, Blink is a collection with more “natural looking” people. A customer looking for “real people,” not professional models, might want to search Blink. However, if you look at Blink, Crush and Ivy all Premium priced Exclusive RM collections it is hard to see any general style distinction between them.  

A problem with this strategy, as best I can determine, is that nowhere on the site is there any definition of the style characteristics of any of these created collections. I’m not sure how customers are expected to know where to find certain types of imagery. This may be a work in progress.

In each imagery category Corbis has exclusive collections and Media Partner collections that are available through many distributors. In some cases these brands are well known, but other Media Partner brands seem to be invented names that no one has ever heard of. For example, the RM Media Partners are: Amana, Architecture, Blend, Canopy, cultura, Food Passionates, Image Source, moodboard, Opal and Photocuisine. Most of these are well known brands, but who are Architecture, Canopy and Opal? Are these real companies that want to hide their identity? Is more than one company represented in each of these brands? If the images in these collections are non-exclusive then where else can they be found?

Below, I have listed the brands in each category. I have separated the Corbis exclusive brands from those supplied by Media Partners. I have also provided a list of the brand offered at each of the four price levels. Readers should note that many of the brands in the price level list are not included in the easier to find Creative and Editorial lists. It is impossible to determine the category that the imagery in these brands falls under. The software development group is working on finding a way to make it easier for customers to search individual collections.

Slotting

Golden said, “We are looking at making some changes in ‘slotting.’ We slot premium and exclusive collections higher because we believe Corbis customers don’t want to see the same images on every site.” However, they slot Blend Images, cultura and Image Source images very high and those collections are with many distributors.

“We do a lot of analysis of the number of times images are previewed, put in lightboxes and purchased and weight the slotting on how customers interact with the content,” Golden continued.

They also want to change the slotting to meet the needs of specific to customer groups. For example collections that are used in North America are different from the collections used in Europe. They are working on delivering different result to these different groups depending on where the customer making the search is located. They expect to beta that project later this fall and go live with the feature in 2012.


IMAGES EXCLUSIVE TO CORBIS

RM	RF	ILLUSTRATION	DOCUMENTARY
Blink	Fancy	Cherry	Corbis Collection
Comet	Flame		Encyclopedia
Crush			Latitude
Ivy			Terra

IMAGES FROM MEDIA PARTNERS

RM	RF	ILLUSTRATION	DOCUMENTARY
Amama	Alloy	Images.com	Nat. Geographic
Architecture	Blend	Motif	Passage
Blend	Collage	Prism
Canopy	cultura
cultura	Image Source
Food Passionate	Juice
Image Source	moodboard
moodboard	Ocean
Opal	Photo Alto
Photocuisine	Radius
	Rubberball

ARCHIVAL	NEWS & SPORTS	FINE ART	GALLARIES
Bettmann	Corbis News	Andy Warhol Foundation	New to the Fine Art Collection
Conde Nast	Corbis Sports	Ansel Adams	Contemporary Works
Historical Collection	Corbis Wire	Blue Lantern Studio	Famous Portraits
Nat. Geo Vintage		Brett Weston Archive	Icons of World Art
Retro	ENTERTAINMENT	Brooklyn Museum	Major Historical Events
Sygma	Corbis Entertainment	Christie's Images	Masters of Impressionism
Turnley	Corbis Wire	Fine Art Photo Library
		Fratelli Alinari
<20 YEARS OLD	OUTLINE	The Gallery Collection
Corbis Entertainment	Outline	Historical Picture Library
Corbis News		Philadelphia Mus. Of Art
Corbis Sports		The Picture Desk Ltd.
Corbis Wire		Smithsonian Institution
Sygma		SWIM Ink
Turnley
VII

PRICING LEVELS

PREMIUM	STANDARD	VALUE	SUPER VALUE
Andy Warhol Foundation	Alloy	Alloy	Boost
Ansel Adams	Amana	Bridge	Design Pics
Bettmann Premium	Amana RF	Collage	Image 100
Blink	arabianEye	Corbis Yellow	Imageshop
Canopy	Architecure RF	Cusp	mingle
Cardinal	Architecture RM	Documentary Value	moodboard
Conde Nast	Asia Images	Flirt	Somos
Corbis Entertainment Premium	Bettmann	Motif	Spirit
Corbis News Premium	Blend RM	Nomad
Crush	Brett Weston Archive	Ocean
Davis Lynn Wildlife	Canopy	PhotoAlto
Fancy	Cherry	Prism
Historical Premium	Christie's	Spirit
Ivy	Collage	Surf
Latitude	Comet	Value Art
Sygma Premium	Contemporary Artists
Terra	Corbis Art
zefa	Corbis Entertainment
	Corbis Museum
	Corbis News
	Corbis Sports
	Corbis Wire
	culture
	cultura RM
	Design Pics
	Documentary
	Encyclopedia
	es
	First Light
	Flame
	Folio
	Food Plassionates
	Fotosearch Illustration
	Fratelli Alinari
	Historical
	Historical Picture Library
	Image Source
	image 100
	Images.com RM
	InsideOutPix
	Juice
	MedicalRF.com
	moodboard
	moodboard plus RM
	Motif
	National Geographic
	National Geographic Vintage
	Ocean
	Passage
	PhotoAlto
	PhotoCuisine
	plainpicture
	Prism
	Radius
	Retro
	Roger Ressmeyer-Starlight
	Ron Chapple RF
	Rubberball
	Schlegelmilch Photography
	Science Photo Library
	Smithsonian Institution
	Solus
	Somos
	Swim Ink
	Sygma
	Terra
	Tetra
	The Gallery Collection
	The Picture Desk Ltd.
	Turnley
	zefa