MicrostockGroup Sponsors


Author Topic: best match Indexing Problem - A Severely Damaging Issue  (Read 5270 times)

0 Members and 1 Guest are viewing this topic.

« on: November 24, 2012, 09:48 »
0
I hadn't realized how broken the best match search really was until I did some test searches on my own today.

I have read threads where people talked about how when you add a second search term that your search results include images that are either brand new or instead rather old pictures but that also have 0 or very few downloads. I see now though that the search engine on iStock is severely broken and its no wonder why buyers aren't buying anymore. It seems they just can't find what they are looking for.

Today I did a test search for the word "train". Then I added the word "elevated" to my existing search results.

The first picture that came up for me provided by the best match search is one that has 0 views and 0 downloads indicating that it has no history and can't possibly be deemed as a best match to my search. 

The worst part of it though is that the picture doesn't contain an elevated train at all, but a train way off in the distance passing through on the ground and in a valley. So not even an elevated train.

The picture itself is titled "Hiking In The Swiss Alps", which further indicates to me that the train is not the main intended subject of the photo by the contributor. Rather hiking in the mountains seems to be the main intended subject of this image.

What follows that picture is a whole slew of at least 2 more rows of other unrelated train pictures with low and/or 0 views and 0 downloads as well.

If I were a buyer I would simply go to another web site to buy my pictures after getting such poorly related search results like that. How can a buyer even be expected to find and buy pictures they need when the search results are not fully related and are coming out like this? It is often already difficult enough to find what you want with so many images matching your keywords, but when the images you are getting aren't even related to your search, let alone ones that have no history of popularity or downloads, what else can you do but go elsewhere?

Apparently the search has been like this for a few months. And what I am afraid of is that if such a severe problem like this, and one that is so directly related to a buyer's ability to buy what they need, hasn't been fixed by now, then it may not get fixed anytime soon. And the site will just continue to lose more and more buyers by the day out of pure frustration.

How this can go on for so long is puzzling and, with no explanation from iStock to buyers about the problem or what they can do to try and find the pictures they need, I don't blame the buyers for going elsewhere. I would too.

Hopefully iStock will be able to fix this problem before its too late, but sadly the prospects of that aren't very promising seeing how everyone's download numbers seem to continue to worsen every month.
« Last Edit: November 24, 2012, 10:16 by bokehgal »


Poncke

« Reply #1 on: November 24, 2012, 10:45 »
0
When I search for elevated train, all I get is images of elevated trains

ShadySue

  • There is a crack in everything
« Reply #2 on: November 24, 2012, 10:58 »
0
If you search on Elevated Train, is your result better? On my search, most of the top best match photos have trains which are on raised up tracks. Elevated Train is in the controlled vocabulary, so is a legitimate search term. (But we are all being served up different results based on our location and buying history, the latter of which is almost non-existent for me.)

Otherwise, you are finding any image with both 'elevated' and 'train' in the photo, which would legitimately include an elevated view of a train on the ground, for example. You would not be so likely to find an elevated train by searching first train then elevated. It's not the way the system works. The file which you mention is in position 16 on my M search. It has a train and perhaps, arguably, is a slightly elevated view, though that's possibly debatable.

It's not unusual to find new images (with no sales) at the top of the best match search. Otherwise, what would the point be of uploading new files. ATM, it seems that there are more old images with no/few sales at the top. The thing is that files can be lost forever if you happen to upload them at a time when new sales are sinking like stones. Recently, I had a newly uploaded file down around positon 230 less than a day after it appeared in my port, so it might never be found. Conversely, this week I've sold quite a few licenses for old files with few/no previous sales, showing that just because a file sank on upload doesn't mean it's worthless. Conversely again, I had three files which recently were licensed within a week of upload which were punished the next day by sinking down below position 300 on their main search term, and other files uploaded before these, but with no sales, are still in the top 50.

However, the search system is badly broken in many other ways. It seems that BM2, which was designed to punish spam, didn't deliver in terms of sales. Or something. But they seem to have abandoned it, and spam is utterly rife, as it is on most sites. For example, "Commercial Kitchen" is currently a far better best match search than it has been for months, but add 'nobody' and you'll see how much spamming goes on. There must have been a huge best match tweak this week, as 'commercial kitchen' as a CV keyword phrase (not in quotes) is cleaner than it's been for years, on my search. Also, earlier this week pawnbroker's best match was unbelievably spammed, confirmed by someone else and now it's pretty good. E+ files are still being hammered, though, as they have been still E+ was introduced.

« Reply #3 on: November 24, 2012, 10:59 »
0
@Poncke - You are correct if you enter the keyword "elevated train" as a single search term.

Enter the keyword "train" by itself. Then enter the keyword "elevated" to your existing search results on train.

The results have changed a bit for me from when I first did the search yesterday to what they are now.

That photo that was in first position, Hiking in The Swiss Alps, has now dropped down to image position 16.

The results are a bit better now in terms of being more related to my search term, but the results are not nearly as good as if you enter "elevated train" as a single search term.

The first image I get now is an editorial image of a train parked in a train station in China with 0 views and 0 downloads. Not expected best match results at all.
« Last Edit: November 24, 2012, 11:14 by bokehgal »

« Reply #4 on: November 24, 2012, 11:13 »
0
@ShadySue - Thanks, what you mentioned about the difference between controlled vocabulary terms and separating these two search terms makes sense in this case. But getting so many new pictures coming up and hardly any good sellers, E+, etc is also a serious problem as you mention. Sure, it will help to sell some new pictures, but if the results aren't mixed enough the buyers still won't be happy.

On the controlled vocabulary search term "elevated train" the first 200 pictures you don't get an E+ file until you get to picture #40. And you don't get your first picture with over 100 downloads until you get to picture #199.

ShadySue

  • There is a crack in everything
« Reply #5 on: November 24, 2012, 11:13 »
+1


The results are a bit better now in terms of being more related to my search term, but the results are not nearly as good as if you enter "elevated train" as a single search term.

The first image I get now is an editorial image of a train parked in a train station in China with 0 views and 0 downloads. Not expected best match results at all.

If you want to see images with many downloads, search by downloads.

If you search 'elevated' and 'train' separately, you'll get all files with the two keywords 'elevated' and 'train'. They don't have to refer to elevated trains. To get an elevated train, you should search on 'elevated train', but you don't need the quotes as 'elevated train' is in the CV as a search phrase.

That is the system working as it should. It's not like, for example Alamy, where if your pseudonym (which might be your real name) is George and you have an image of a square, it will show up in a search for George Square.

ShadySue

  • There is a crack in everything
« Reply #6 on: November 24, 2012, 11:15 »
0
The first image I get now is an editorial image of a train parked in a train station in China with 0 views and 0 downloads. Not expected best match results at all.
You can opt not to see editorial images if they wouldn't suit your intended use.

That train is a sky train (position 22 in my search), i.e. designed as an 'elevated train', even though it isn't 'elevated' in that actual photo. So arguably not technically incorrect.
« Last Edit: November 24, 2012, 11:22 by ShadySue »

« Reply #7 on: November 24, 2012, 11:21 »
0
@ShadySue - Thanks. But it is the quality of the mix on this search which as I mentioned is also the problem even on the CV word "elevated train". Sure, sorting by DL will give you all the top sellers, but the art of the best match was it used to give a mix of pictures across the range of categories, like top sellers, E+, pictures with high views, recently sold pictures, etc. Now you get mainly new pictures with few downloads and/or views. Only showing those pictures means the results aren't going to pop out at buyers. That's a problem.

As I said on elevated train, you get your first E+ at 40 and your first 100+ image at 199.
« Last Edit: November 24, 2012, 11:25 by bokehgal »

ShadySue

  • There is a crack in everything
« Reply #8 on: November 24, 2012, 11:33 »
0
That's very new. For over six weeks, new files have been dropping very quickly. I'm glad they fixed that. There isn't much motivation to upload new files if they're off the first page of 200 within a day of being uploaded, evn if there are only 250 files or so altogether on that search. I agree it seems odd to penalise a file for selling quickly by dropping it below the first page.

It'll change again very soon, no doubt. It always does. There is no doubt no way to achieve a perfect balance, especially with all the spam.

Also in this new search, which is no more than three days old, they seem to have fixed the non-CV phrase search. For a long time, non-CV phrases were being split, even if searched within quotes. So if you search non CV phrases as "Word1 Word2" you're supposed to get that. For weeks now, even if you searched within quotes, the two words were split in the search. Now it seems to be working, at least with two-word searches.

There also seems to be a person bias, and not the way you might imagine. I don't want to get personal by pointing out examples. But there's one person, who has very few overall sales, who has flooded some 'obscure' searches in which I feature. Although in most of the searches his many near-identical files have 0 dls, he's hogging the top several lines, and has done for months.

tab62

« Reply #9 on: November 24, 2012, 11:36 »
0
Okay, I am somewhat confused (probably born this way) on the way you are using the search engine? I want to Shutter and typed in fish- it brought back over 276K plus images. I than typed in rainbow to see if it would narrow my search to Rainbow Trout. I did not! Instead it brought up images of rainbows. Am I missing something?

Poncke

« Reply #10 on: November 24, 2012, 11:42 »
0
So basically there is no best match Indexing Problem - A Severely Damaging Issue

tab62

« Reply #11 on: November 24, 2012, 11:46 »
0
I took some programming courses, years ago, and the search engine doesn't go into sub-sets as mention above. You have to put all your criterion in the initial search to get what you want. Google works the same way and I consider them the industry standard...

« Reply #12 on: November 24, 2012, 12:20 »
0
Yes, the best match is broken. Many searches give you really poor results and it's unbelievable that they haven't fixed it yet.

It's been confirmed by an administrator in the iStock forum: "The issue is still being actively explored and yes, there is definitely something wrong. Once we get more information, we should be able to provide an update."

tab62

« Reply #13 on: November 24, 2012, 13:23 »
0
more buyers going to Shutter if that is the case...

ShadySue

  • There is a crack in everything
« Reply #14 on: November 24, 2012, 13:36 »
0
Yes, the best match is broken. Many searches give you really poor results and it's unbelievable that they haven't fixed it yet.

It's been confirmed by an administrator in the iStock forum: "The issue is still being actively explored and yes, there is definitely something wrong. Once we get more information, we should be able to provide an update."

There has been a huge change in the best match since that was written, in the last 2ish days. I'm not saying all the problems have been fixed, but most of my usual searches have changed considerably in that timeframe.

ShadySue

  • There is a crack in everything
« Reply #15 on: November 24, 2012, 14:17 »
0
Okay, I am somewhat confused (probably born this way) on the way you are using the search engine? I want to Shutter and typed in fish- it brought back over 276K plus images. I than typed in rainbow to see if it would narrow my search to Rainbow Trout. I did not! Instead it brought up images of rainbows. Am I missing something?
I have no idea how SS search works. I know it doesn't work with a CV. If I wanted Rainbow Trout, that's what I'd search on. Under the iS CV you shouldn't tag Rainbow in a photo of a rainbow trout except in the extremely unlikely event that it would also have a rainbow in the photo. You would tag 'rainbow' and 'rainbow trout' as well as 'fish'.
Alamy's IMO inferior search engine would presumaly present rainbow trouts if you searched for rainbow, as it doesn't recognise keyword phrases.
« Last Edit: November 24, 2012, 15:50 by ShadySue »

« Reply #16 on: November 24, 2012, 14:23 »
0
Alamy's IMO inferior search engine [...] doesn't recognise keyword phrases.

Huh?  But they let you put phrases in quotes.  You mean they don't actually use them?

ShadySue

  • There is a crack in everything
« Reply #17 on: November 24, 2012, 15:15 »
+1
Alamy's IMO inferior search engine [...] doesn't recognise keyword phrases.

Huh?  But they let you put phrases in quotes.  You mean they don't actually use them?
Nope. It's well known and it never has worked.

« Reply #18 on: November 24, 2012, 15:36 »
0
Yes, the best match is broken. Many searches give you really poor results and it's unbelievable that they haven't fixed it yet.

It's been confirmed by an administrator in the iStock forum: "The issue is still being actively explored and yes, there is definitely something wrong. Once we get more information, we should be able to provide an update."

There has been a huge change in the best match since that was written, in the last 2ish days. I'm not saying all the problems have been fixed, but most of my usual searches have changed considerably in that timeframe.


Ok, that's good to hear. I hope they'll have it back to normal very soon.

« Reply #19 on: November 24, 2012, 16:30 »
0
If things have been fixed in terms of faulty search results then that's great news and we should see some movement in sales from that hopefully.

But the type of dominant file types coming up in best match, predominantly files with very low downloads and/or 0 views and 0 DL is very disconcerting.

We all know the mix has to be balanced with a cross section to give both new files and good selling files and equal chance to sell and gain popularity, but at the moment the mix is not evenly weighted at all and it has been this way for quite some time I've noticed.

Other than the search results not satisfying the buyers in one way or another, what else can account for the 30-50% drop in sales that many contributors have been experiencing?

Unless of course there has simply been a large mass exodus of buyers from iStock.

ShadySue

  • There is a crack in everything
« Reply #20 on: November 24, 2012, 19:11 »
0
If things have been fixed in terms of faulty search results then that's great news and we should see some movement in sales from that hopefully.

I didn't say it was totally fixed. A search for Sunderbans brings up many thousands of photos, only a few related to the Sunderbans - embarrassingly including five of mine.
Maybe Sunderbans means something in one of the 'community languages'.

KB

« Reply #21 on: November 24, 2012, 23:18 »
0
A search for Sunderbans brings up many thousands of photos, only a few related to the Sunderbans - embarrassingly including five of mine.
Maybe Sunderbans means something in one of the 'community languages'.
I don't think they consider that behavior a bug, even though most sane people might. As 'Sunderbans' isn't in the CV, it decides to look for what it thinks are close matches and displays those. Often the results are worthless, but every so often it probably guesses right.

KB

« Reply #22 on: November 24, 2012, 23:20 »
0
Alamy's IMO inferior search engine [...] doesn't recognise keyword phrases.

Huh?  But they let you put phrases in quotes.  You mean they don't actually use them?
Nope. It's well known and it never has worked.
If I type South Dakota into the search box, I get 12,440 results.

If I type "South Dakota" in, I get 12,073 results.

It's trying to do something with quotes, but who knows what in the world that is.

Anyway, sorry for the detour, this is OT for this thread.

gillian vann

  • *Gillian*
« Reply #23 on: November 25, 2012, 00:46 »
0
is it normal to put in 3 of your keywords into the search and not see your file come up?

« Reply #24 on: November 25, 2012, 03:32 »
0
is it normal to put in 3 of your keywords into the search and not see your file come up?
I rarely go to IS  but if I do searches for any reason I have also had that happen.  I'm suprised that they have any buyers left at all it is such a frustrating site.

Poncke

« Reply #25 on: November 25, 2012, 04:30 »
0
If things have been fixed in terms of faulty search results then that's great news and we should see some movement in sales from that hopefully.

I didn't say it was totally fixed. A search for Sunderbans brings up many thousands of photos, only a few related to the Sunderbans - embarrassingly including five of mine.
Maybe Sunderbans means something in one of the 'community languages'.
Its spelled Sundarbans, maybe that makes a difference?

aspp

« Reply #26 on: November 25, 2012, 05:19 »
0
Alamy's IMO inferior search engine [...] doesn't recognise keyword phrases.

Huh?  But they let you put phrases in quotes.  You mean they don't actually use them?
Nope. It's well known and it never has worked.
If I type South Dakota into the search box, I get 12,440 results.

If I type "South Dakota" in, I get 12,073 results.

It's trying to do something with quotes, but who knows what in the world that is.

Without quotes it brings up all images which have south and dakota anywhere in the keywords. When you use quotes it brings up only images which have the two words adjacent to each other in that exact order.

For example - there are 9 images of C47 Dakota aircraft which also have "south" as a keyword. These will show in a search without the quotes.
« Last Edit: November 25, 2012, 05:21 by aspp »

ShadySue

  • There is a crack in everything
« Reply #27 on: November 25, 2012, 08:35 »
0
A search for Sunderbans brings up many thousands of photos, only a few related to the Sunderbans - embarrassingly including five of mine.
Maybe Sunderbans means something in one of the 'community languages'.
I don't think they consider that behavior a bug, even though most sane people might. As 'Sunderbans' isn't in the CV, it decides to look for what it thinks are close matches and displays those. Often the results are worthless, but every so often it probably guesses right.

If it's not a bug, it's a bif (badly implemented feature).
(Ooooh, I just made up a techy acronym  :D)
It is quite good at making a reasonable guess if you make a typo in the Search Box, but if they are making a guess, they should tell you what they're giving you, as in 'did you mean 'Sunderland'? (that's not it, I doubt if any I see were taken in Sunderland any more than the Sunderbans). Otherwise, a searcher has no idea whether they might get relevant results somewhere or there are no hits and they should try elsewhere. I don't think anyone wanting a photo of the Sunderbans wants bikini-clad models on random beaches, which many results (not mine) showed.


 

Sponsors

Mega Bundle of 5,900+ Professional Lightroom Presets

Microstock Poll Results

Sponsors