After publishing my analysis of PicturEngine last week (9/14) Justin Brinson, PicturEngine CEO, made extensive comments. I’ve decided to re-publish the entire story with Justin’s comments inserted in italics and indented where he indicated. I hope this gives readers a clearer understanding of this new search engine.
After years of development
PicturEngine is about ready to begin promoting its search engine to customers. It has more than 500,000,000 keyword searchable images from 64 stock photo agencies and hundreds—soon to be thousands—of individual photographers.
The site is designed to put control of image distribution and commissions back into the hands of photographers where PicturEngine believes it rightfully belongs.
The site is designed to eliminate duplicates when the same image is with multiple agencies. If the photographer’s prime agency has put the image with several sub-agencies, buyers will be directed to the prime agency for licensing, and the images that are in the collections of other sites will not be shown. If the photographer has placed an image with PicturEngine directly and the same image is on multiple other sites non-exclusively licensing will be handled by the photographer directly and the same image will not be shown as being represented by any of the other agencies.
The theory behind this is to, as much as possible, either let the photographer control the licensing directly, or to designate one prime agency to handle all negotiation, and eliminate the multiple sub-agent cuts that are so prevalent in the industry today.
When customers search on the site, there is no indication as to the number of images found in the search. Unlike most other sites, there are no page breaks requiring the user to click to go to the next page of 100 or 200 thumbnails. Instead, the user just keeps scrolling down – in theory through tens-of-thousands of images – never knowing when they might be getting near the end.
Some of our results appear in real-time, so it is not practical to show a count that continues to grow as you scroll down and interact. As you scroll, we are adding, sorting, and ranking in real-time as you interact with our search.
Our “more like this" feature is disabled at the moment that will give you even more change as you scroll. Organic results load first, which are surpassing 700 million images before removing exact duplicates. Results grow with every search performed on the platform. We are currently ranking “organically" so sometimes sites with a lot of images from their collection of a specific subject appear higher in the results than our Platform or Advertising photographers/agencies. I believe that if we rank ALL of our PicturEngine Platform and Advertising photographers/agencies first, then we cannot say the search is “unbiased." Instead, we give the user what we think they are looking for. With more traffic, tracking clicks and interactions, we will be able to zero in on this more, as each search helps us build our ranking system. There is a local (individual user) and global component to our search. Soon with more users searching, each individual user's results will be different, as we find out specifically what each user likes to search for, view, add to lightboxes, and license.
As you scroll, we are adding in our “Advertising” images in sets, showing the buyer sets of images from both "organic" and "paid advertising” searches. When we tried evenly distributed results (think shuffling cards) it looked odd with sims from the same photographer separated, which appeared as duplicates. Is it perfect? No, but is it improving every day with more searching? YES! The more a buyer searches, adds to lightboxes, and clicks on a particular subject, the more we can zero in on what that particular buyer is looking for!
If the searcher clicks on a preview, it is shown with all of its information under the last row of thumbnails. To continue searching the user simply scrolls down below the preview and the next row of thumbnails is shown.
I did a search for London and 20 of the first 25 pictures shown are Wolfgang Kaehler’s pictures of the London Eye. This is not necessarily a good cross section of photographers, agencies, or the general subject matter available to someone looking for London. I narrowed the search to “London Big Ben” and got a lot of Getty Images pictures, but not a lot of images from other sources, at least initially. In addition, in many of the first images, while Big Ben was there, it was a very minor and insignificant part of the image.
All commonly used Boolean search operators and search options (AND, OR, NOT, -people, -person) work within our search. Users are familiar with using these on many search sites. Here is a link I found on Google:) http://library.uaf.edu/ls101-boolean
We also support special operators for our industry. You can place RM or RF after a term (or search string) to return just those license types.
In short, search filters are either already within our search or coming. We just need to build the GUI for those that don’t know to use them. I am not a fan of clutter or buttons. Starting off we are trying to keep it as SIMPLE as possible to get people searching and exploring. It’s getting better every day and we are learning and adapting to search habits.
For instance, a search for 'London' should bring up all images with the word London, ranked by specialty (if a collection is known for a certain subject those images get an initial boost). In your example, you saw our search in action; a smaller collection with a lot of “London” photos got a boost. HOWEVER, if that collection is seen and is not favored by buyers, you will see that keyword specialty boost begin to diminish. If buyers like it, it will remain the same. If buyers really like it and add these images to a lot of lightboxes or click to license those images, then those specific images and the overall collection, gets more of a boost. I try not to talk about the inner working of our search results ranking algorithms too much, as we don’t want users trying to game the system. Then we will need to counteract or filter their influence on our overall ranking. These measures were meticulously calculated, planned, and tested based up what we learned from my previous agencies' searches and search experience with actual image buyers, as well as countermeasures for photographers trying to game the system to get higher ranking of their images.
I hold over 10 years of saved search history and habits from Cornerhousestock.com and Picturesque.com, HOWEVER, little of the current traffic we are receiving (specifically from posts about PicturEngine to photographers) follows those models of image buyers. My speculation on our current search traffic is that thousands of photographers and niche agencies, are searching to find out if we have their images listed in our index. They want to see their current image placement, so they search for a specific “keyword” that they KNOW is unique to their images. We understand that this traffic is not buyer traffic, so we are recording, tagging, and labeling it accordingly.
I searched for “United States Capitol.” This should have given me pictures of the U.S. Capitol building. The first images were mostly credited to Getty Image. If there were several shots from one photographer, they seemed to be lumped together in one sequence. The 30th image in the sequence was a shot by John Elk III of the Capitol building in Minnesota. After the initial Getty photos there were a long series of Depositphoto images and then a return to more Getty Images.??
At about thumbnail 150 there was a series of about 20 frames taken by Tim McGuire of the Capitol Dome from slightly varied angles. Tim is a direct contributor to Picturengine, but all of his images are superseded by a lot of agency images.
I searched for “Egypt” and again got a lot of Getty Images pictures in the beginning. One 537022793 of a Greater Sand Plover was taken by “David Verdonck/NIS/Minden Pictures - Getty Images.” It seems that Getty will do the licensing of that image and there will be three cuts of the fee before David receives any money.
We try to determine the base agency whenever possible. This process is semi automated at the moment and becoming better and more automated every week. It would fix everything perfectly if all of the smaller base agencies (that pay the photographers directly) would sign on and become advertisers, then we could filter everything perfectly, removing all sub agencies overnight! Until then, we will do what we can to send the buyer as close to the seller (or creator) as possible.
Side by side in the Egypt search, there were two exact same images credited to Peter Phipp, Getty Images. These images had two different Getty Images numbers – 160833017 and 527950185 – so evidently Getty has included the same image twice in its collection. To narrow my search, I added Sphinx (“Egypt Sphinx”) to my search term. Not too far down in the search I found two Fotolia images. One (32112628) was of Sphinx above the Opera House in Budapest was taken by Viadislav Gajic and the other (90140717) was a Sphinx on the embankment of Neva River in St. Petersburg taken by Anna Pakutina. If a researcher is looking for the Sphinx in Egypt these aren’t too useful.
Visual search is not turned on yet, so visual deduping is not running. We found that not only duplicate content exists at separate or multiple agencies, but a lot of agencies have duplicate content within their own collections. This duplication is a real problem in the industry and we are going to solve it with our search.
I searched for “Monument Valley.” Among the first pictures shown was a Getty Images picture 556443071 credited to Danita Delimont. That image will be licensed by Getty Images and a royalty paid to Danita Delimont Stock Photography. I suspect the image was actually taken by one of the photographers Danita represents. That photographer will receive a share of what Danita receives from Getty. Also among the first pictures was 80361400 credited to Tim Fitzharris/Minden Pictures – Getty Images. In addition, there was a Joe Sohm pictures 520375200 of Saddles on a Hitching Post in Monument Valley credited to Getty Images. This image will be licensed by Getty and a percentage paid to Sohm, but Joe Sohm is one of the photographers who has put his images on Picturengine directly -- in theory eliminating that agency cut.
It is possible that Danita Delimont Stock Photography and Minden Pictures are not included among the 64 agencies currently participating on PicturEngine. Agencies with at least 1 million keyword searchable images, and online pricing and delivery, are included on the site for free. Smaller agencies can join the archive for a small monthly fee.
PicturEngine makes its money to cover costs and profits by charging photographers and smaller agencies a small monthly fee for making their images available through the site. To review the pricing schedule see
here.
I like to remind people that not everyone uses (pays for) Google Adwords, or Bing Ads, to draw traffic to their website, but these advertisements pay for the rest of us to use these search engines for free. It is a model that WORKS!
Keys To Success
1 – To be successful PicturEngine must make it easier for image buyers to search all the images of a particular subject that are on multiple sites. If the images on
GettyImages.com are always the first to show up, why should the buyers bother? Just go to Getty.
Getty has some great images, so we include all of them. Some may have better ranking than others. More searching will show you that no single agency dominates our search. If that were to become a problem, we will adjust our search accordingly.
Why go to Getty or any agency as a first stop when PicturEngine has a larger view of the whole industry? For the FIRST TIME, image buyers can organize, share, and collaborate on images in lightboxes across the WHOLE industry. Did I mention our buyer tools are unique and awesome?!
2 – Images of photographers who place their image with PicturEngine directly must come up first. Images from agencies that have been designated as the photographer’s primary agency must come up second. Images from the agencies where the image creators have made no designation must follow the other two. That doesn’t seem to be happening.
I believe it's better to view a mix and let the buyers decide what they want to see. Google recently received scrutiny for only showing their paid advertisers in their top shopping tool. We want to be the BEST search for stock photos and not just another agency.
3 – Customers must be able to quickly review a broad cross section of the images being shown in a particular keyword category. Given the number of similars shot by the same photographer of the same situation that show up side by side, that does not seem to be happening. Over time PictureEngine will be able to determine which images are clicked most frequently (by customers, not photographers looking at their own images) and show them first. However, getting to that point requires a lot of customer interactions. In the meantime, customers aren’t getting much benefit from the huge database of images.
Visual sim stacking (for individual photographer sims) is coming. We need more interactions from actual image buyers to tell us what image we should place on the front of the stack. The image that grabs the buyer's attention is not always obvious. We want to get the buyers to click to see what is included as sims, but first we would like them to tell us what they like.
Here is an example of sim stacking in action on a PicturEngine Platform license page: http://www.picturengine.com/stock_photo_detail/view/image-Mur025641569766.html
When sim stacking is used in the search results, it is not the big box tall as it is on the license page (there is no room for that) instead we are experimenting with a side scrolling box and also something that resembles apples' cover flow view (although we have found that cover flow is very resource intensive). More testing is needed.
4 – Getting a significant number of customers to use a new site is not easy. It takes a huge amount of marketing. To get customers to switch from the current site they are using, they must see that somehow it will be a huge benefit to them. It’s hard to see where the benefit is for customers today.
Our search is improving every day. The more buyers use our search engine, lightbox, and tools, the better our results. I believe it will be easy for buyers to discover our benefits of a better way to search for stock photos.
5 – One of the problems customers face is that they have no idea, until they open the preview, whether an image is RM or RF or the approximate cost of licensing the image. Customers with budget requirements could spend a lot more time trying to find an image that fits their budget. It is probably more practical to go to sites that offer prices they understand.
You can search NOW for just RM or RF by placing RM or RF after your search string. GUI functionality is coming. Try: London Big Ben RM or London Big Ben RF
In addition, increasingly customers that spend the most money on photography want to negotiate a discounted price-per-image used, plus special services and legal indemnities if they only use images from that particular agency. PicturEngine has no way of effectively servicing those customers.
If a buyer has a special pricing “deal” with a particular agency, they will likely not use PicturEngine.
However, when buyers discover that the PicturEngine Platform images, have a “Make Offer” system built into in the pricing of Platform images, putting them in direct contact with the photographer that holds copyright and SHOT the image, they will see a HUGE benefit and I believe a game changer for our industry.
Example: Buyer sees an image they like. They price the image for a license and click "Make Offer" instead of “Add to Cart." The buyer supplies a little more detail of the intended use and submits the offer. We instantly notify the photographer, letting them know they have a pending offer. The photographer then has one of three options:
1. "Accept" the offer at the price and terms the buyer specifies. In this case, the buyer is instantly sent an "Offer Accepted" notification and can checkout and instantly receive the image, at the terms and price they specified with all details written to the license.
2. “Decline" the offer. In this case, the image price reverts back to the original set price by the photographer in the pricing calculator in the buyer's cart. The buyer is sent an instant notification that their offer was declined, then can make the decision to license or not.
3. The photographer can submit a "Counter Offer," updating photographer notes and the price, and add the new counter offer to the cart. Both buyer and seller's notes are included in the license, including the original price, offered price, and countered price.
This system was initially implemented to supplement any deficits in our pricing calculator (we wanted to keep our calculations easy for BOTH buyers and sellers), but when we realized that buyers love it, we expanded it and will continue to improve the concept as it matures. We essentially automated the negotiation process, which buyers love for PicturEngine Platform Photographers, and put negotiations in the photographer's hands, and in realtime. Everyone has a smartphone or email to respond to offers quickly, so why not automate it? As of today we have email and instant site notifications, and we are testing SMS text notifications if the user chooses this method. Of course we are also working on an app with push notifications for both buyer and seller. We built into our PicturEngine Platform system ease-of-use, ease of negotiating licenses, transparency, and FULL documentation of licenses. ALL offers and notes are documented and written on each image license. This is just the beginning! We have so many tools built into our PicturEngine platform and system, to help photographers price, keyword, and get their images in front of buyers as fast as possible, so that they are competitive in this industry. PicturEngine is the future of the stock photo industry.
NOTE: After I JUST explained our offer system and all the hard work and years of testing and trials we put into it, I am sure it will be copied by agencies, BUT our system is still unique where buyers are negotiating directly with photographers that receive 100% of the sale. I don’t think agencies will do that!
Generally I stepped back, studied and learned form the most successful business models in the world, before and while building PicturEngine.
Google/Bing: Our search engine advertising model component (without the rampant image theft and price comparison)
Amazon: Group of small, medium, and large sellers using the same platform with one checkout that buyers LOVE
eBay: Direct buyer and seller interaction/negotiation with a governing platform to assist both
We are thinking outside the box, implementing concepts from what these undisputedly successful business models have done, and put our own spin on it for our stock photo industry.
PicturEngine is released to the public with a few features disabled. To turn on all the bells and whistles, visual search, More Like This, real-time search deduping, phone and tablet front-end formatting, and more, PicturEngine must first get itself to market.
Please note that search results at
http://www.picturengine.com/ will fluctuate over the next few weeks as they prepare for their public campaign and bring new data online. They take photographer suggestions very seriously, and want to ensure that everything runs smoothly. If you discover an issue, or have feedback of any kind, please send a message to
fhccbeg@cvpgheratvar.pbz. Updated FAQs are available at
http://support.picturengine.com/
One last thing you touched on, but I want to emphasize. We really LOVE suggestions and feedback, even complaints are fine! Send them our way! The culmination of feedback collected from image buyers, photographers (being a photographer myself), and managing/owning agencies myself over the last 20 years, created PicturEngine.
We built what the industry asked for, and we sincerely appreciate your continued support.
To learn more about the platform features for photographers check out this
link.