Photographic forums have been dominated over the last week by discussion of Alamy’s new search algorithms, which have changed the order in which search results are presented to a buyer on the photographic library’s internet site.
Specialist subject contributors, for whom Alamy had become an important part of their business, found that images which had previously been returned at the top of search results were now being returned only on later pages, which are far less likely to result in a sale. Photographers spoken to by EPUK say such a change in the search results could mean a drop in royalties of a thousand pounds a month
>> Privately owned UK-based stock agency established in 2001
>> Popular among both professional and amateur photographers for no entry charges, and is open to all who meet its technical requirements.
>> Offers royalties of up to 65% on sales, well above the industry average
>> Has currently over 7 million images on file, up from 4 million a year ago
However, Alamy CEO James West told EPUK: “I would also be cautious about judging how the system is behaving based on test searches that are thought to be representative of customer behaviour. The only way to objectively measure the impact of AlamyRank will be after enough time has passed to build a realistic picture of what is happening. Of Alamy’s seven million images, can anyone really say how many belong to specialists and how many of those have been adversely affected?”
“Sudden drop” in sales
His comments are unlikely to reassure specialist photographers and libraries. In various photographic discussion forums, including EPUK’s discussion list, many are reporting a sudden drop in prominence, with badly or irrelevantly keyworded images heading the search results
Many specialists also hold images on their subject taken by other photographers. Alamy calls these subsets “pseudonyms” and the specialists found that these pseudonyms were far better placed than their own work, an effect which some photographers feel can only be explained by AlamyRank having penalised large collections compared to smaller sets.
Alamy’s policy of allowing contributors to keyword and edit their own images has led to criticism from buyers that search results can produce large amounts of very similar images before finding anything useful, and that many images are excessively keyworded with irrelevant terms.
In response to this, Alamy’s solution to make search results more relevant is based around taking into account the past “success rate” of each contributor. The patented system, named “AlamyRank”, works by assigning each contributor a rank built up from their past performance over the last few months.
Every time a picture is viewed but ignored by a buyer the photographer or agency’s rank is reduced by a tiny fraction, whereas when a buyer shows interest in an image by clicking on it, the contributor’s rank is boosted by a rather larger fraction. So contributors who fill the pages with near identical images or who appear in every search because they have – in the words of Alamy CEO James West “stuffed the dictionary into the file” – lose rank compared to more disciplined photographers.
How AlamyRank works
“The new search order is controlled by three components: the search engine, AlamyRank, and the Diversity Algorithm.” Alamy CEO James West told EPUK. “The search engine works in exactly the same way as before AlamyRank and is concerned only with returning images that are relevant to what the customer types in the search box.” Alamy say that there is no manual intervention into the system which is purely based on automatic analysis of buyer behaviour.
Alamy’s James West illustrates example AlamyRank calculations at a recent meeting with contributors Photograph: Pete Jenkins
“AlamyRank is used to score collections of images according to the number of times images from a collection (defined as all the images that are associated with a pseudonym) have been clicked or purchased in proportion to the number of times they have been seen.”
“This is currently based on 9 months of customer activity and will be updated at intervals to be determined by what we learn in the coming weeks.”
The final element, the Diversity Algorithm, shuffles the order of the images. West told EPUK: “We first order images by the highest combined score for relevancy and rank. The diversity algorithm ensures that the images in the results are mixed up rather than dominated by the contributor with the highest combined score.”
AlamyRank “will educate buyers”
West told EPUK: “Customers on average look at 2,000 images per purchase. That’s a lot of pages of thumbnails. It may turn out to be the case that AlamyRank will, in time, educate customers that they don’t need to drill so far down in the results. In reality I think that for the system to manifest itself as a dramatic improvement for the customer, it must first be much more interactive with our contributors so that the overall quality of submissions and metadata improves over time in response to what AlamyRank is telling them about their pictures.”
This does little to comfort those who are seeing their Alamy income, built up over several years of fitting their workflow to Alamy’s needs, drop by hundreds or even thousands of pounds a month. It is difficult to reconcile Alamy’s long term failure to provide effective tools for contributors to manage their large image collections with their desire for contributors to be more interactive and improve the quality of their submissions. Many experienced specialist stock professionals with the agency are now doubting whether Alamy has a real commitment to their work.
EPUK is keen to hear from any Alamy contributors who have been affected by the Alamy search algorithm change – contact us at the editor’s email address below left
Want to contact the EPUK Website editor? firstname.lastname@example.org