MicrostockGroup Sponsors


Author Topic: Someone tricked the Shutterstock algorithm?  (Read 33583 times)

0 Members and 1 Guest are viewing this topic.

Justanotherphotographer

« Reply #25 on: April 26, 2021, 03:15 »
+1
His linkd profile does exist. You need to search Google for his profile picture. This takes you to a contact. He is listed in that persons contacts on linkd. Listed as his current position are these details.

QA Test Lead AS PNB Banka
AS PNB Banka
Nov 2012 - Present8 years 6 months

Latvia

- Organize the work of the testing department
- automatic test creation (Java, Selenium, JUnit)
- Control the quality of products developed
- to provide a technical task of developing
- authorization device (DigiPass) Storage and Release.
- Internet bank and the bank's Web page maintenance / administration.
- The update preparation to the real environment (Internet banks and the bank's web-site, SVN, GIT)
- Incident solving.


Source https://tinyurl.com/27sthyp7
Okay, so could be programming to farm clicks or certain amount of dls from subs accounts to boost positions


« Reply #26 on: April 26, 2021, 06:47 »
+5
Okay, so could be programming to farm clicks or certain amount of dls from subs accounts to boost positions

If that's what he did, then why only Shutterstock and not on other agencies? He has 53 sales on Dreamstime, which is on the normal or even lower side for a portfolio of that.
I wonder how much it would cost, how much of your own images you must buy, to reach the front page with so many images for highly competitive keywords.
It would also require a lot of subscription plans owned by different companies or persons to pull it off. If that's what happened, this would mean there's a shady industry behind it also offering similar services to other contributors. Meaning: you can buy content ranking. 

Or... it might indeed be a weird Shutterstock experiment as well. See how buyers react on quality and image ranking and how it compares to other agencies.
Push a regular image to the front page and see how long it's able to hold position.

Hell I don't know. But for sure it's manipulated. We can all agree to that.


OM

« Reply #27 on: April 27, 2021, 05:49 »
+3
I would doubt that this is any kind of SS experiment. If you search 'man', you get the top 2 rows on Page 1 'relevant' almost completely filled with his shots. All editorial and not (IMHO) hugely attractive. SS states that it wants model-released shots if possible so why would they go out of their way to present buyers with the exact opposite of what they tell contributors that sells. All social media and Google changed their algorithms a few years ago because some clever guys/gals were able to manipulate the results to their own advantage. Now it's much more difficult. Maybe SS didn't see the need and this guy is smarter than their old algo......or maybe he's just a hacker and SS don't really care anyway!

« Reply #28 on: April 27, 2021, 07:55 »
0
...
Interesting theory, but since he has real accounts on various stock agencies, I doubt that's what's going on here.
Yeah, thanks, so not an SS owned portfolio
Cannot be a final conclusion. Nothing stops one agency to (ab)use another one, no matter under which sauce - AI persona or a real person, and still all other versions are too alive to be eliminated :-)

Uncle Pete

  • Great Place by a Great Lake - My Home Port
« Reply #29 on: April 27, 2021, 10:46 »
0

Hell I don't know. But for sure it's manipulated. We can all agree to that.

Years ago "Red One" had some artificial boost on video and everything that had that as the camera, flew to the front. I guess that his camera or something in the user ID and info has a featured word or programming error. If we can find out and everyone add that, SS will eventually fix what's broken.  :)

« Reply #30 on: April 27, 2021, 16:10 »
+6
I tried something simple to check if it is the keywords or the title:
(a) Original image from OlegRi: https://www.shutterstock.com/de/image-photo/red-white-rose-on-black-background-1589147812
(b) My image submitted with the exact same keywords and title: https://www.shutterstock.com/de/image-photo/red-white-rose-on-black-background-1961404534

Results:
(a) OlegRi's images is very high if you look for flowers or rose
(b) My image cannot be found within the first thousands of images

So its not the keywords nor the title...

« Reply #31 on: April 28, 2021, 01:30 »
+1
Huge thank you to people who can dedicate a part of their time for such research. Stock photographers and other creatives are much less protected category than other businesses.

Justanotherphotographer

« Reply #32 on: April 28, 2021, 02:50 »
0
Okay, so could be programming to farm clicks or certain amount of dls from subs accounts to boost positions

If that's what he did, then why only Shutterstock and not on other agencies? ...

Could be he saw the bigger earnings potential so concentrated on beating the algorithm at SS.
Could be that it does involve purchasing a certain number of images and the maths just makes sense on SS for investment/ return

OM

« Reply #33 on: April 28, 2021, 04:44 »
+2
I think that if they and multiple friends were buying the images, they would have to buy an awful lot to get his images as the first 10 images on page 1 of 'man'....a category of 22 million images. Same goes for all the other most popular categories in which they have one or more shots in the top slots. You would have to purchase a lot of photos.

In the old days when there were far fewer images in the SS collection I had a couple of images (almost) at the top of search in popular, single-search-word categories and for that those images were selling between 10 and 20X/week and every week for months in order to stay at the top.

« Reply #34 on: April 28, 2021, 14:44 »
+3
I don't believe the theory that he / his friends / someone he pays is buying loads of these images, either. This is not like paying a click farm 50 bucks once to get 5.000 fake followers for your company's social media channels. He would have to buy all these images constantly, and that would cost a good deal of money over time. But the way his search results look, he would just be wasting a lot of money.

Take the keyword "man" as an example. I counted 34 of his images on the first page of results. As OM mentioned, they are all editorial, so they are useless for commercial customers.

Also, some are a bad fit even for customers looking for editorial. For example: people in Venetian masks, a person with a metal detector on a beach who has is back turned to the camera, a pseudo-old-timey statue that is holding up a glass of beer, and loads of people on motorbikes - none of those are what I would want if I typed "man" into the search bar.

So I just don't think those pictures sell very well, no matter how high up they are in the search.

Someone who is bumping his images to the top by buying them repeatedly would concentrate on a few that have commercial value. At least, no one would bother bumping up the beer glass statue guy - that's just like throwing money into a burning pit.
« Last Edit: April 28, 2021, 15:43 by lazy_ella »

Uncle Pete

  • Great Place by a Great Lake - My Home Port
« Reply #35 on: April 29, 2021, 11:56 »
0

So I just don't think those pictures sell very well, no matter how high up they are in the search.

Someone who is bumping his images to the top by buying them repeatedly would concentrate on a few that have commercial value. At least, no one would bother bumping up the beer glass statue guy - that's just like throwing money into a burning pit.

Thank you, that's what I was thinking also. First page doesn't mean sales, if the images aren't good.

But I sure wonder how he's doing it?

Is the SS algorithm so stupid that a click farm can raise someones presence in the searches enough to make them first page? That's very unusual to imagine it's so easy to manipulate. And that would seem to indicate it's by artist not by individual images. Mine don't appear to display that way, they are by image and sales.

How sick would that be if anyone could raise their level and visibility just by having many more views?  :o



« Reply #36 on: April 29, 2021, 13:00 »
0

How sick would that be if anyone could raise their level and visibility just by having many more views?  :o
Kind of DDOS?

« Reply #37 on: April 30, 2021, 02:46 »
0

Thank you, that's what I was thinking also. First page doesn't mean sales, if the images aren't good.

But I sure wonder how he's doing it?

Is the SS algorithm so stupid that a click farm can raise someones presence in the searches enough to make them first page? That's very unusual to imagine it's so easy to manipulate. And that would seem to indicate it's by artist not by individual images. Mine don't appear to display that way, they are by image and sales.

How sick would that be if anyone could raise their level and visibility just by having many more views?  :o

Far from being the expert here, but I would assume that CTR (Click Through Rate) is indeed one of the factors that determines the ranking of an image by the algorithm. Conversion rate (actual licensing) should be another one. In other words: if an image gets a lot of clicks we can assume it rises up in the ranks, but if nobody is buying it after closer inspection, it probably drops again.

I discovered yesterday that SS seems to be implementing these factors also in their front end. Popularity and Usage are on the image details page. At least, on my phone. Not (yet) on my PC. One of my images that get licensed several times a week since I uploaded it last year has low popularity and usage and an image that sold a few times more than a year ago and never again since then has high popularity and usage.

I would also assume that SS is able to detect basic manipulations. Clicks coming from the same account, PC, IP address or region should be discarded.

« Reply #38 on: April 30, 2021, 03:05 »
0
Roscoe, "Clicks coming from the same account, PC, IP address or region should be discarded." -
They will react only when sales can be affected. It is not a problem to forge many attributes, for example, Yahoo is not able to stop bitcoin ads spam, everytime it is coming from different addresses etc.

« Reply #39 on: April 30, 2021, 03:13 »
+1
They will react only when sales can be affected. It is not a problem to forge many attributes, for example, Yahoo is not able to stop bitcoin ads spam, everytime it is coming from different addresses etc.

Yes, that's why I said "I would assume". Not sure they have this fraud detection built in.

Should be easy to test, no? Let's pick a random low quality image from deep down the SS haystack which we find by using a certain keyword, start clicking it, and see if it rises up the ranks.

« Reply #40 on: May 01, 2021, 07:35 »
+1
SS aren't likely to care about this much.

For them, a sale is a sale.  Who they pay the tiny commission too doesn't matter.

I suspect they care very little about any potential algorithm abuse as long as its not affecting sales.

Horizon

    This user is banned.
« Reply #41 on: May 01, 2021, 10:06 »
0
Hope you all realize that we as contributors see a different search then the registered buyer!!...old news I know but still!...if I was to search using a certain keyword which is highly popular I have 7 pics on the first page and they would of course sell like crazy but they dont!!  not even once!....why? because in reallity and with the buyers search the pics are probably way down the line!


« Reply #42 on: May 01, 2021, 10:36 »
+1
I don't think buying your own images through click farms would make any sense financially, especially considering the subs image prices.
I think it is views that take precedence in a lot of cases, not sales.

Here is the example - I have uploaded the image 3 months ago. I had 2 sales only - 10 cents each, but the image is located in the first place in the first row. So it cannot be sales that influence the position in this case. The search was performed without being logged in to my contributor account, so it is the way clients see it.
'Teotihuacan' is not a hugely competitive keyword, but still a popular tourist destination in Mexico.

Back to the original point. If views take precedence then it is much easier to write the script to dupe the system than to deal with a complicated procedure of producing fake sales.

« Reply #43 on: May 01, 2021, 12:29 »
0
It seems the latest plausible theory is that these people spam their images to the max after acceptance, wait for them to pick up speed and when they find a place in the top results they remove the spam to appear regular. Apparently many users do this. Look at these:
https://www.shutterstock.com/image-photo/mountain-waterfall-landscape-mountains-1780718075
https://www.shutterstock.com/image-photo/winter-mountain-snow-camp-scene-camping-1780718078

« Reply #44 on: May 01, 2021, 13:28 »
0
Hope you all realize that we as contributors see a different search then the registered buyer!!...old news I know but still!...if I was to search using a certain keyword which is highly popular I have 7 pics on the first page and they would of course sell like crazy but they dont!!  not even once!....why? because in reallity and with the buyers search the pics are probably way down the line!

I have a customer account and I see the same problems

« Reply #45 on: May 01, 2021, 14:16 »
0
Hope you all realize that we as contributors see a different search then the registered buyer!!...old news I know but still!...if I was to search using a certain keyword which is highly popular I have 7 pics on the first page and they would of course sell like crazy but they dont!!  not even once!....why? because in reallity and with the buyers search the pics are probably way down the line!

I have a customer account and I see the same problems
Same here. Customer & Contributor account search gives me the same results...

« Reply #46 on: May 01, 2021, 16:40 »
+1
It seems the latest plausible theory is that these people spam their images to the max after acceptance, wait for them to pick up speed and when they find a place in the top results they remove the spam to appear regular. Apparently many users do this. Look at these:
https://www.shutterstock.com/image-photo/mountain-waterfall-landscape-mountains-1780718075
https://www.shutterstock.com/image-photo/winter-mountain-snow-camp-scene-camping-1780718078

You'd guess the second one uploaded then as soon as it got approved keyword spammed it.  I cant see that number of duplicate keywords getting through even SS's dumb-bot QA process.


« Reply #47 on: May 02, 2021, 00:06 »
+1
I've seen contrubutors do this before, but as far as I can tell, no other contributor could recreate Olge's "success" the way he did.
« Last Edit: May 02, 2021, 00:15 by Firn »

« Reply #48 on: May 03, 2021, 07:49 »
+2

https://www.shutterstock.com/image-photo/winter-mountain-snow-camp-scene-camping-1780718078

On a side note: the second one (Andrey Bocharov) does not only keyword spamming, the photos in his port are stolen, too. Found some in other portfolios, he just flipped them. Too bad Shittersuck wouldn't care if I contacted them.

« Reply #49 on: May 03, 2021, 11:43 »
+1

https://www.shutterstock.com/image-photo/winter-mountain-snow-camp-scene-camping-1780718078

On a side note: the second one (Andrey Bocharov) does not only keyword spamming, the photos in his port are stolen, too. Found some in other portfolios, he just flipped them. Too bad Shittersuck wouldn't care if I contacted them.

Good spot - appears he's stolen this guys off Flickr:-
https://www.flickr.com/photos/cheafmundl/49652681852/

Just flipped it and increased the saturation.  Its definitely the same image (look at the clouds).

So he's not just gaming the system and keyword spamming, he's doing so with stolen images on top.


 

Related Topics

  Subject / Started by Replies Last post
4 Replies
4276 Views
Last post March 15, 2016, 19:20
by michaeldb
45 Replies
23814 Views
Last post May 05, 2016, 12:08
by Minsc
71 Replies
30567 Views
Last post February 04, 2018, 10:45
by YadaYadaYada
34 Replies
19083 Views
Last post November 07, 2018, 18:41
by farbled
20 Replies
8738 Views
Last post September 29, 2018, 15:59
by nobody

Sponsors

Mega Bundle of 5,900+ Professional Lightroom Presets

Microstock Poll Results

Sponsors