Quote from: Neurobite on November 26, 2018, 15:30

Quote from: Not Today on November 26, 2018, 14:52

Quote from: Neurobite on November 26, 2018, 13:35

Quote from: MatHayward on November 26, 2018, 11:10

Quote from: Pauws99 on November 26, 2018, 11:01

I was responding to this

"everytime buyers search for images and come up with irrelevant results your images get marked down as less relevant" - sorry, do you have an evidence of such thing???"

It is safe to assume that if an image appears in a certain search result, is seen by the customer and not purchased that the image drops lower in the search results. It is best to only add keywords that are relevant to the image. Irrelevant keywords may be considered spam and have a negative effect on both the potential of success for your file and on the customer experience.

-Mat

...Thanks Mat for jumping into this discussion; only you can clarify things which are in mind of AS contributors... However, even good assumptions do not build strong bridges.. I still have a few unclear for me points:

1) Does AS algorithm really punish ("drops lower in the search results") good images because they were viewed but not purchased??? Do you know that any buyer (!!!) before buying an image selects 10 best ones, carefully looks at them, then selects 3 from 10, and then goes to sleep, next morning, he/she has to do a hard task, to select 1 from 3 and buy it... Though, in other words, AS assigns less weights to those poor 9 images, because they were viewed but not purchased? Do I understand it correctly?

2) What does it have to do with bad, spam, fishing keywording? You name it... Because, if a buyer is looking for an apple, but a page with 100 results shows him/her 1 peach, in 99% the buyer even will not look at it, no clicks, no views, you have no data to work with, and the algorithm has no idea, why peach was not viewed, is that because bad keywords, or because some other 100+ parameters: colors, background, composition, crop... In other words, it is extremely difficult task to look into "black matter" of buyers mentality and predict their behavior...

3) I think the best way, and, probably, the only way, to fight spammers and fishers is to use an image recognition system, an AI (there are a lot of them already in the market), and delete the most irrelevant keywords at the stage of submitting your content... SS is already doing this... I think AS is already using a kind of this artificial intelligence (AI) software, because it can recognize duplicates or very similar content in your portfolio and reject them for submission... The same should be done for keywords!!!

1) I personally purchase photos and videos regularly for video editing purpose for my clients, and I don't have the luxury to go to sleep and select what I want the next day or the day after... because deadline. Usually if I see something I like in the first pages, I just buy it and get on with my work. Sometimes I might need approval for specific projects so I'll copy the links and send them to the client so he can choose the ones he likes best, but definitely I'm not going to pick something because it's pretty and not relevant.

2) View means appear in search results. So your 'peach' will go down in ranking if it keeps appearing for 'apple' search in the search results but that is not the only factor taken in consideration in the algorithm, that would be too easy to cheat the system, it's a mix of sales, clicks and views, among many other things.

3) I wouldn't trust AI softwares at this point, it's not accurate enough and can't always recognise properly what's in the image let alone video.

...yes, that's what I meant, between searching, selection and purchasing there are many things happening, some go to sleep, some show content to clients or project leaders, some buy instantly cause they have money, so they buy whatever they like for their own database for 100+ projects in mind... However, what I wanted to stress here is that buyers do not buy content based on how well keywords match to their search strategy/expectations... they buy on basis of the particular image likeness, attractiveness, technical and aesthetic quality, and how well it fits to the project...

...your second point I did not get: how on earth algorithm knows that there is a peach actually in the image but not an apple and it should not be in the search results? The peach with "apple" in keywords will be shown in the "apple" search results, but machine thinks it is an apple .. and there will be 10000+ more unseen, unviewed unclicked truly apples, by your hypothesis, they should be all punished, because they were in search results but not viewed... unless there is image recognition which will tell the script the peach is not matching to the apple, let's punish only the peach... Do you understand what I mean?

...well, if you never tried Google image/place recognition service, I suggest to try it, you will be astonished how accurate it is...

Let's say if 100 people search for 'apple' and nobody clicks on or buys your 'peach' image, your image becomes irrelevant, and goes down in ranking. All the other images of apples in the same situation will also be 'punished' if you want to call it that - they are not relevant in the sense that nobody wants them, so they will also go down in rankings.

Back to your Waterfall image for instance, every time someone looks for actual photos of Sea or Flowers as the main subject, and scroll past your photo without clicking on it or buying it, your photo will go down in ranking. So it's a gamble, you might attract a random buyer who's in love with your photo while they were actually looking for something else, but in the meantime, every time someone isn't in love with it, your photo will go down.. until it's buried under photos of other contributors. Is it worth the risk?

...well, I am not sure that AS script works this way, I will explain what I mean: your algorithms with dynamic ranking (can go up and down, can be rewarded and punished) is too difficult to implement and run, and actually, no need for that... I think there is only cumulative historical scoring for each image, scripts operate with clear-cut variables, such as: number of sells, number of views, ratio views/sells, number of collections the image is put, date (age of the image)... that's it... you submitted a new image it acquires scoring, lets say 1, each its sell adds to it 1 point, each view 0.1, put in collection 0.01... your scoring can only grow up, based on image performance... Then, lets think how image search engine works? - you search for "apple", you have 1M results of images which have "apple" in their keywords, then the script select first 100 images with highest scoring, and makes first result page for you, then next page, the script may cook differently, it will show you more fresh images still with highest scoring, 3 page might have another layout, it will select best scored images from the middle... and so on... Many companies do such cocktails to maximize chance to sell for all images, not only highest ranked... Your imaginary punishment in ranking or scoring occurs naturally, passively, images which gain points push others backward... I think scoring images down by whatever reasons in active way is a fairy tail... Unless Mat says yes, we have it...

I'm sorry neurobite but you had the explanation from Mat who works with Adobe.

You know you are keyword spamming, we know it and the buyers will definitely know it.

You are just writing lengthy posts to justify it.

Out of interest I looked for that waterfall with the extra keyword "flowers" and one page of results but only two images contained anything remotely like flowers.

Yours had no flowers so I did not click on it and neither did I save it to my light box which is exactly the behaviour you could expect from a buyer.

I have some spare subscription credits left so I think I'll go buy the two images that did contain flowers.

Alamy tell us this

" Pseudonyms with a high click through rate (CTR) will be receiving more zooms relative to views than pseudonyms with a low CTR. This is important information because the ratio of zooms to views is one of the factors that influences your AlamyRank score.

Is CTR my AlamyRank?

No. CTR is just one of the factors that make up your AlamyRank."

I think its a very reasonable assumption that many if not most agencies include this to some extent in their search algorithms.

Quote from: Pauws99 on November 27, 2018, 04:26

Alamy tell us this

" Pseudonyms with a high click through rate (CTR) will be receiving more zooms relative to views than pseudonyms with a low CTR. This is important information because the ratio of zooms to views is one of the factors that influences your AlamyRank score.

Is CTR my AlamyRank?

No. CTR is just one of the factors that make up your AlamyRank."

I think its a very reasonable assumption that many if not most agencies include this to some extent in their search algorithms.

Yeah Alamy takes it even further as this CTR score affects your whole portfolio/pseudo's image rankings and not just the image itself. Wondering if it's the same with Adobe Stock, would many low ranking images have an impact on your portfolio as a whole?

Its really anyone's guess even though some claim special knowledge...we do know adobe collect zoom stats so why wouldn't they? My guess is that sites have a "contributor ranking" based on a number of metrics that feeds into search algorithm.