236 posts Marketplace Developer
  • Has collected 1+ items on Envato Market
  • Member of the Envato Team
  • Located in New Zealand
  • Has been part of the Envato Community for over 2 years
madlep Envato team says

Couple of fixes deployed today:

  • “Element 3D” Videohive nav link fixed. This one wasn’t updated to point at the new URL, but at the old one. There is a redirect that forwards to the new one, but it mangled the attribute value, so no results showed up when the search on the attribute key was done.
  • Synonym list pruned to remove many redundant synonyms that were causing too many items to be recalled e.g. searching by “race” on videohive, and sorting by sales had items with very low relevancy included. This is a simpler fix that could be done quickly. A better solution will be to cut off items with a low relevancy score, and just not include them at all.
  • Justin Bieber made special guest appearance at Envato HQ, and live coded new search features while performing his greatest hits.

Other stuff:


all these test/improvements are great. can you tell us when this testing period is done so we can actually grow in sales if we publish more items instead of publish more items just to lose less percentage of our income each month?

Can you identify specific items or categories that have taken a hit? If you can provide specific info, we can investigate, and potentially make changes if there are problems. I’m having a look at your sales count and earnings data, and I’m not easily able to see exactly what the issue is…


I am selling less and less videos, as it seems… And with 1300+ footage files in my portfolio (and still growing…) it is a statistical improbability… these days I usually get sales for videos which have already been bought a few times in the past, or at least that’s my impression…

Again, if you can provide specific examples, it would be very helpful. Hard to pinpoint the exact issue from your sales and earnings data…


Imho, the whole point of new search functionality was to provide buyers to discover files they need, not to provide yet another place where already popular files will be promoted and those less popular to get buried even more. I am commenting like this because so far I’ve read in a few occasions that sales are (or are going to be) used to boost search rankings… Why would it matter? If it does matter, buyers should be able to sort by sales and that’s it, why use sales as a default parameter to boost search ranking…?

Sales count is a strong indicator of quality, which is why it’s one factor affecting search relevancy score. Sorting by sales has issues (as we’ve seen), in that it completely disregards how well an items matches your query, and just returns EVERYTHING with even a weak match to the term.


The same applies to new VS old items – if needed, there should be a sort functionality. I know it might be important for ThemeForest, CodeCanyon, or for AE templates (where newer items use more modern technologies/techniques) but what about music, sound effects, photos, footage…? Does the upload really matter? It can only make things worse imho. Different marketplace segments should be approached differently, there just can’t be the same pattern for everything anymore… Or am I wrong? Am I missing something?

Definitely each marketplace needs it’s own needs, and requires it’s own approach to balancing search. This is something we’ll be looking at.


what I want is for a clear date when the search had been updated, what item description guidelines we need to follow.

There is no “updated date”. It’s an ongoing incremental process.


If planned search system features are secret because competition may steal the great ideas developers have, fine by me.

We’re not worried about anyone stealing our “great ideas” (we’re not really doing anything that cutting edge). Any information we hold back is to avoid authors from gaming items to get to the top of search results for queries that are totally irrelevant to their content. That doesn’t help authors or buyers in the long run.


I don’t like the continuous improvement plan. This means continuous playing with search and with my nerves.

This is the way envato always has, and always will work. It’s nothing new.

We’ve always had a strong automated test suite to support working this way. However, we’ve been lacking that for search relevancy testing.

What we are going to be doing is investing more in automatically testing search results of a new release against vs production so we can immediately highlight if and where search results have been affected by a change. The approach is to replay the top 1000 queries from each site, and report differences in ordering for the top few hundred items.

We did some of this with the initial elasticsearch prototype rollout (more info here), but we didn’t maintain that tooling as we rolled out search. That is something we should have done.

5180 posts
  • Elite Author: Sold more than $75,000 on Envato Market
  • Has sold $125,000+ on Envato Market
  • Helps us moderate the forums
  • Provided great contribution to our forums
+12 more
felt_tips Moderator says

stuff…

Description used too heavily in the search results will by its very nature lead to erroneous results, and to the temptation for authors to stuff their descriptions with keywords. I guess at the very least links, embedded images and so on could be filtered out.

Regarding the synonyms, is it possible to use a more complex and nuanced synonym matching? Rather than simply putting in a couple of synonyms to a word, to use some kind of existent synonym database to decide how closely words are related to one another? I’m guessing they exist and are what’s behind those word trees you sometimes see on thesaurus websites. That might avoid some of the circumstances that Diego describes above, where a word is given synonyms which are not synonymous, but simply loosely related and fairly arbitrarily chosen terms.

I don’t see how “race” is remotely synonymous with “car”. A car is an object, a race is an event, which on some occasions is an event involving cars, but could equally well involve horses, athletes, dogs, rats, bicycles, the clock or any number of things. And that’s before you consider “race” in the sense of “human race”. Are “athlete” and “horse” synonyms for “race” too? Is this going to have to be done with every search term? How on earth were these synonyms thought up? Is it not obvious that having “car” as a synonym for “race” (but not “horse”, “athlete”, “clock” etc.) deeply biases the results? What about other related terms? “Lap”, “track”, “time” etc.

If synonym search is going to be used, then there needs to be a clear separation between different forms (e.g. “race”, “races”, “racing”, “raced”), true synonyms (“contest”, “competition”, “sprint” etc.) and related terms (e.g. “car”, “track”, “lap”, “horse”, “prize”). The true synonyms are in themselves already tricky, because actually, “derby” and “grand prix” are synonyms for race, but far more specific ones than “contest”. And as for related terms, because this is also not a binary, related/not-related relationship, but a continuum … and indeed one that is very subjective.

@randomnoise… I totally agree (not just for Audiojungle, but especially for Audiojungle) that the search can’t be truly solid until categories, tags and meta-data are seriously restructured.

no-thumbnail
accudatamarketing Recent Posts Threads Started
13 posts
  • Has collected 10+ items on Envato Market
accudatamarketing says

Perhaps you should take duplicate content into consideration for search results. That way if authors are copying a popular files content and tags it gets placed below the fold or something. I think that would help out a lot because many times we come across items that are of low quality but the copy looks like it was taken from another item with a few changes in order to exploit the search results.

Just a thought!

1412 posts
  • Has sold $10,000+ on Envato Market
  • Member of the Envato Team
  • Made it to the Authors' Hall of Fame
  • Had an item featured on Envato Market
+9 more
lucafrancini Envato team says

I think Randomnoise wrote some interesting suggestions about AJ search, and I hope these are things that will be considered for the future, to make audio search easier and more effective for buyers

From my humble experience as “recurring uploader” I’ve noticed an increase in sales/exposure on my new items, since the engine has changed. Months ago they made one or two sales and then disappeared from search results and had drop in views/sales; now both views and sales are steadier 10 – 15 days after the upload. Maybe this is thanks to the item’s sales/date relation or whatever; anyway I’m happy, because this way new items have more chances to become popular. :)

7606 posts
  • Has been part of the Envato Community for over 5 years
  • Has sold $40,000+ on Envato Market
  • Sells items exclusively on Envato Market
  • Located in Europe
+5 more
doru says

Can you identify specific items or categories that have taken a hit? If you can provide specific info, we can investigate, and potentially make changes if there are problems. I’m having a look at your sales count and earnings data, and I’m not easily able to see exactly what the issue is….

quote from my dashboard page:

” Fun fact

Your last sale was $6.00 1 day ago “

Writing here that you’ve checked and then saying there is no problem is puzzling and I’m starting to think you are doing it on purpose.(the same as how you keep continue removing my comments from this thread – already did this two times)

Today is 4 March and there are 4 sales. One of that is off course the 6$ dollar item above.

On 4 February I’ve had 9 sales. February 1 and 2 was a weekend,

On 4 January I’ve had 11 sales.

On 4 December – 10 sales

On 4 November – 12 sales

This looks to me as not in line with the previous months

181 posts
  • Located in Brazil
  • Has sold $10,000+ on Envato Market
  • Helped several times protecting Envato Market against copyright violations
  • Has collected 50+ items on Envato Market
+5 more
DiegoMirnov says

Hi Madlep, another example: type: “engine”, sort by sales. A real mess will appear. I wanna search all these items related to engine(machine) and lots of files appear, even with no relations. I already know in some(not all) of these projects there is the word “engine”, and some with unfair advantage. The first place in the list is using lots of keywords between commas, like: ” search engine optimization”. 3 words, not one. Videohive rule: put max 15 keywords. In older days, you couldn´t put lots of words in each keyword, cause you´ll probably lose visibility. And now, you don´t need so much the keywords help, cause if you write “engine” in item description, your file will be on the list. So, that bad idea i told: “put the dictionary in item description could make a simple project appear in various searches” could work.

P.S.: type: “engine”, sort by sales. There is a project called “Code Awake” that don´t have the “engine” word. It has “engineering” word.

236 posts Marketplace Developer
  • Has collected 1+ items on Envato Market
  • Member of the Envato Team
  • Located in New Zealand
  • Has been part of the Envato Community for over 2 years
madlep Envato team says

(lots of great ideas!)

All good suggestions. We’ve got a few of those on our list, but I’ll make sure they get noted.

Yes, faceted search is definitely being worked on now in the design/analysis phase. It’s still a while off, but it’s getting there.


Description used too heavily in the search results will by its very nature lead to erroneous results, and to the temptation for authors to stuff their descriptions with keywords. I guess at the very least links, embedded images and so on could be filtered out.

Yup, definite risk. We don’t give description a high boosting, and this is handled to some extent by the way TF-IDF works to naturally favour shorter documents that contain a term more prominently – so keyword stuffing tends to drown itself out. That all goes out the window when sorting by sales though, which is something we’re still working on fixing


Regarding the synonyms, is it possible to use a more complex and nuanced synonym matching? Rather than simply putting in a couple of synonyms to a word, to use some kind of existent synonym database to decide how closely words are related to one another? I’m guessing they exist and are what’s behind those word trees you sometimes see on thesaurus websites. That might avoid some of the circumstances that Diego describes above, where a word is given synonyms which are not synonymous, but simply loosely related and fairly arbitrarily chosen terms.

I don’t see how “race” is remotely synonymous with “car”. A car is an object, a race is an event, which on some occasions is an event involving cars, but could equally well involve horses, athletes, dogs, rats, bicycles, the clock or any number of things. And that’s before you consider “race” in the sense of “human race”. Are “athlete” and “horse” synonyms for “race” too? Is this going to have to be done with every search term? How on earth were these synonyms thought up? Is it not obvious that having “car” as a synonym for “race” (but not “horse”, “athlete”, “clock” etc.) deeply biases the results? What about other related terms? “Lap”, “track”, “time” etc.

If synonym search is going to be used, then there needs to be a clear separation between different forms (e.g. “race”, “races”, “racing”, “raced”), true synonyms (“contest”, “competition”, “sprint” etc.) and related terms (e.g. “car”, “track”, “lap”, “horse”, “prize”). The true synonyms are in themselves already tricky, because actually, “derby” and “grand prix” are synonyms for race, but far more specific ones than “contest”. And as for related terms, because this is also not a binary, related/not-related relationship, but a continuum … and indeed one that is very subjective.

The synonym support in lucene/elasticsearch isn’t that sophisticated unfortunately. You can map one token for another, or for multiple others, but it doesn’t have a built in concept of synonym strength.

Something we want to go deeper into is building smarter “suggested searches” based on synonyms, rather than just using them to expand the result set. Suggesting “car” as an alternative search when you search for “race” makes sense, but I totally agree that “race” probably shouldn’t be expanded to “car” for search results.


Writing here that you’ve checked and then saying there is no problem is puzzling and I’m starting to think you are doing it on purpose.(the same as how you keep continue removing my comments from this thread – already did this two times)

I’ve sent a separate message outside of the forum, about sales/earnings, as I don’t want to discuss potentially sensitive sales info here.

Your other posts were removed as they were not contributing to the purpose of this thread – which is to collect specific actionable feedback that the dev team can use to improve search.


Hi Madlep, another example: type: “engine”, sort by sales. A real mess will appear. I wanna search all these items related to engine(machine) and lots of files appear, even with no relations. I already know in some(not all) of these projects there is the word “engine”, and some with unfair advantage. The first place in the list is using lots of keywords between commas, like: ” search engine optimization”. 3 words, not one. Videohive rule: put max 15 keywords. In older days, you couldn´t put lots of words in each keyword, cause you´ll probably lose visibility. And now, you don´t need so much the keywords help, cause if you write “engine” in item description, your file will be on the list. So, that bad idea i told: “put the dictionary in item description could make a simple project appear in various searches” could work. P.S.: type: “engine”, sort by sales. There is a project called “Code Awake” that don´t have the “engine” word. It has “engineering” word.

Yup, that’s a mess. Sort by sales is definitely problematic. We’ve got a few fixes we’re working on to make it behave better. A start would be to filter out items that don’t match a certain relevancy score threshold.

As for engine/engineering – that’s how stemming works, and it probably makes sense for most cases. Usually, it would result in a low relevancy score though, unless it was used somewhere prominently in the item (like in the title etc)

5180 posts
  • Elite Author: Sold more than $75,000 on Envato Market
  • Has sold $125,000+ on Envato Market
  • Helps us moderate the forums
  • Provided great contribution to our forums
+12 more
felt_tips Moderator says


Description used too heavily in the search results will by its very nature lead to erroneous results, and to the temptation for authors to stuff their descriptions with keywords. I guess at the very least links, embedded images and so on could be filtered out.
Yup, definite risk. We don’t give description a high boosting, and this is handled to some extent by the way TF-IDF works to naturally favour shorter documents that contain a term more prominently – so keyword stuffing tends to drown itself out. That all goes out the window when sorting by sales though, which is something we’re still working on fixing

Understood – so the “the”, “and” and “but” bits get sorted out, and its about keyword density in the document.

But nevertheless, as long as the description is used at all in the search, a good strategy for the author is to filter out all terms that are not likely to be searched upon and maximize the density of good search terms for his/her item (more or less the way one goes about thinking up tags, actually)

The danger is though, that a lot of information that is very important to buyers is not search friendly. Not only is it not search friendly, it dilutes the density and the search engine score of the search friendly terms.

Look at my description on this item http://videohive.net/item/sketch/91631

Would you concede that it would rank higher on the terms “sketch”, “preset”, “drawn”, “cool”, “animation” and “look” if I were to remove everything except the first sentence? What I should do is encapsulate the latter part of the description in a gif image and embed that in the description, leaving my searchable keywords as a greater constituent element of the overall description.

You see how search is bending the purpose of the description? As long as this is the case, authors will be writing descriptions that are primarily about SEO and not about giving the customer the most useful and relevant information.

My suggestion would be to stop searching the description altogether and add some more keyword options. These are pure, targeted search tags. Otherwise, as I see it, you’re designing misuse into the system.

31 posts
  • Has referred 1+ members
  • Has sold $10,000+ on Envato Market
  • Has collected 10+ items on Envato Market
  • Located in Europe
+2 more
Themes1 says

Greetings madlep,

Just to add to the ‘items that have taken a hit’ list; we’ve seen a noticeable drop on our end. We normally have a few sales a day, and that since the first item we published here (a year ago)—but now it’s been a week with no sales at all, this definitely has do to with something search-related.

Please advice if you have any suggestions. Thanks.

7606 posts
  • Has been part of the Envato Community for over 5 years
  • Has sold $40,000+ on Envato Market
  • Sells items exclusively on Envato Market
  • Located in Europe
+5 more
doru says

just a question, how much time it takes for the search to get in account your description after you’ve edited?

by
by
by
by
by
by