Can Technology Replace Human Curators?

Everyone agrees there is an oversupply of images. In spite of this fact many professional image buyers claim they can’t find good images or at least the images they need. As I look at what is available online today I think there are more good and great images than there ever have been, but often they are buried under piles of mundane images and images that are irrelevant to buyers needs. The problem is curation.

A Little History

Back in the 80s when a buyer wanted an image they called a picture researcher at a stock agency, described their project and their ideas of what they were looking for, and asked the researcher to send them a selection of images to consider. The researcher was very familiar with what was in the agency’s files. The researcher also knew what other people with the same request had used. In some cases the researcher also had ideas of her own of what might be good for this customer and she put the whole thing into a package that was sent to the customer for consideration.



Often the researcher would go through hundreds if not thousands of images, but would narrow the selection down to a couple hundred at most that she believed were on target for the particular client and project. Buyers tended to call the same researcher for every new request and through conversation, over time, the researcher would get a good understanding of what each buyer wanted.

The agencies accepted lots of images into their files just in case someone might come along in the future wanting that particular subject, but the vast majority of these images were never shown to any buyer because there were better images in the files. Occasionally, a buyer might say, “Send me everything you’ve got on a particular subject.” In this case the buyer was prepared to spend a lot of time looking at everything available, hoping to get a new inspiration, but such requests were rare.

I remember one time when a researcher who had worked for Black Star for years retired. Stock sales fell off rather dramatically for a period of time until the new researchers built relationships with the customers and learned the files. A significant amount of institutional memory had been lost when the woman retired.



In the 90s the industry moved to the print catalog era. Researchers and sales associates would go through their files and select the best images in their collection from every subject category. On average a catalog would have 2,000 images, but some had as many as 6,000. There were maybe 40 to 60 images in each category so it was very easy for an art director to open catalog, find the category of interest for the particular project they were working on and quickly review the agency’s best offering. Some images were brand new and some had been in the files for several years. The selection was based on what the agency thought their customers were most likely to want to use.

It turned out that very high percentage of the images most agencies licensed were ones that could be found in one of their catalogs. Customers also kept the catalogs and use reference them year after year. Major advertising agencies and design firms would collect photo catalogs (they were all free) from every source they could find. When they needed a picture they would go through several catalogs until they found something they could use. Usually they had preferred catalogs that they would go through first.



Some large organizations had libraries dedicated to storing print catalogs with as many as 400 to 500 different books. Agencies like The Image Bank would distribute as many as 300,000 copies of each print catalog they produced to art directors around the world. The agencies had millions of images in their collections – both Getty and Corbis were claiming to have over 70,000,000 images in their collections. Only a very small percentage of their total collections ever made it into catalogs but catalog images represented a majority of the sales.

All this happened when files were curated.

Moving To Digital

In the digital environment the idea was to put everything online and let customers do the search on their own. Initially, when film had to be scanned there was still a lot of curation. But, with digital capture it became easier to put almost everything delivered to the agency online line thus providing customers with more choice.

There continued to be some editing, but in recent years editing has become more and more about whether the image is clean, sharp, has a proper model release and sufficient keywords rather than whether the image is anything customers might actually be willing to pay money to use. A very high percentage of the images delivered by creators go into the online offering.

Today, most of the editors who choose the images to upload have never had a conversation with an image buyer. I doubt if any of them are being provided with much information about what is really selling and what is in demand. They may be given such information in a very generic category way, but when it comes to the images they accept they add everything that meets basic quality standards and fits into some category in the collection.

In the early years of digital delivery customers used to be able to call an agency and say, “I’m working on this project. I’m really busy and I don’t have time to do a thorough image search. Would you put together a light box on the subject and send it to me?” That way the customer could get the benefit of curation and the agency’s expertise. As far as I know almost no one requests that service anymore.

Many of the smaller specialized agencies still offer curation. They have researchers who know their collections and the people who do the research actually talk to customers. They have a better understanding of their subject matter and of their customer’s needs. They will do light boxes. But, unfortunately, because the bigger portals have become so dominant the smaller ones don’t get many requests.

Often the customers who say they can’t find what they need in the big databases don’t know that many of these small specialized agencies exist.  

Keywording

Another problem with image search is keywording. Initially when we moved from print catalogs to digital the agencies were doing the keywording. They, at least, had the opportunity to know what words the customers were using to describe what they needed and to apply those words to the images coming in that fit that profile. They may not have used this opportunity very effectively, but at least it was there.

Now, the image creator, or someone he hires, does the keywording. Usually, that person knows nothing about the buyers or the words they use to describe images. Often they will have no understanding of the photographers thinking or reason for creating the image. This breakdown in communication between customer and creator is why curation is needed.

Solutions

To better serve their customers portals might offer curated collections of images in particular categories. This would be similar to the old print catalogs, but would be searchable online. The number of images in any of these categories should be kept to no more than 200 to 300. However, each category would need to be curated by someone who knows the subject matter and has some knowledge of what customers have been using. Some great images might stay in a particular curated collection for years. But new images added to the overall collection would need to be constantly reviewed to see if they should be made part of the curated collection.

This could help customers find the good images that are now buried and save them a lot of time looking through a bunch of images that are irrelevant to their needs. The problem, and why it is unlikely to happen at current price points, is that in order to implement such an idea there would be a huge additional staff cost.

Microstock

In some senses microstock and subscription have the best current answers to this problem. The “Best Match” search clearly doesn’t solve customer’s problems. A lot of images delivered near the top of this search seem unlikely to ever be used. Search by “Downloads” lets customers see the images that have been most popular with other buyer and this seems to be the default search order preferred by many customers. In effect, the other buyers are providing a degree of curation for the portal.

However, there are two problems with “download” search. Once an image has been on the site for a certain period of time, and received a certain number of uses it will always be pushed to the top of the search return order. This means that images added more recently tend to be lower down and even when they get a significant number of downloads it is hard to get them above the 300th or 400th image returned.

Also, a good image with lots of downloads and 30 or 40 keywords will come up at the top of the search returns when any of those keywords are used, despite the fact that most customers who licensed use of the image only used one or two of those keywords to find it.

If the customer want to check out the new image she can do a “Newest” search, but then she is overwhelmed with images whose only merit is that they were just uploaded. If there are a few good images in the batch, within a couple weeks they may be pushed so far down in the search return that no one who does a “newest” search will ever see them.

However, having these options is much better than the single option on most of the traditional sites that offer RM and traditionally priced RF to customers. On these sites even the best selling images are licensed infrequently compared to what happens on the microstock sites. Thus, very few traditional images are licensed enough for them to standout in an “images downloaded” or “images licensed” search.

Subscription

Subscription sellers have a major advantage when it comes to gathering data because they encourage customers to download images they “like,” but will not necessarily use. Thus, the subscription sites get a lot more data points about a customer’s interests than a site that only tracks actual images licensed.

The other sites could ask customers to click on images the “like” in order to better understand a particular customer’s needs, but I doubt that many customers would go to this trouble unless it could be demonstrated that this greatly improved the other searches they make.
 

Can Technology Replace Curation?

The hope of technologists is that somehow by collecting enough “Data” about the keywords individual customers use as well as the characteristics of the images themselves they can improve curation without the expense of hiring people to do the work.

A big part of the problem is that the only data these online sites get from customers is the keywords they use and the images they buy. They miss all that information that is communicated in personal contact. When it comes to the images, there is no way to tell which of 100 or 1,000 images with all the same keywords will best fulfill a particular customer’s needs. Someone has to look at them. Either there has to be curation, or the customer has to do the looking.?
I think customers are telling us is that they don’t have time to look through all the images we’re offering them and they want some help. Unfortunately, they don’t want to pay any more for this service than they are paying now. They probably want to pay less.

Nevertheless, the data we are currently able to collect is insufficient to make a judgment about what a specific customer will want in the future and particularly insufficient to make a judgment about what another customer who uses the same keyword will want.

Data can help customers search for images that have an overall blue, yellow or salmon color or background, but it is hard to describe in words the mood or feel that the customer gets from looking at an image he/she has never seen. In seems unlikely that technology will be able to breakdown the characteristics of an image in such detail that customers would be able to input such characteristics and narrow their searches to find the images they want.