1617 posts I create things I wish existed
  • Elite Author: Sold more than $75,000 on Envato Market
  • Has sold $75,000+ on Envato Market and is now an Elite Author
  • Won a competition
  • Made it to the Authors' Hall of Fame
+8 more
cyzer says

IMPORTANT

Another search bug/mistake is: if you search for a “scary” it gives you 309 results. But if you want to filter it out by adding two search words “blue scary” it should give only projects which are blue and scary… but now it gives 3101 results….

Every word written in the search acts like a different separate search and stacks results in the same page.

Buyer instead of filtering by adding more tag words actualy makes it harder to find. This is sooooooo wrong.

189 posts
  • Has been part of the Envato Community for over 3 years
  • Has collected 100+ items on Envato Market
  • Sells items exclusively on Envato Market
  • Located in United States
FastNOC says

That’s not really accurate. Adding words to a generic search ADDS variables, it doesn’t isolate them.

the same thing will happen in google. Unless you specify to make your search include BOTH of those words.

a typical search will search for any of the words searched. not all of them

189 posts
  • Has been part of the Envato Community for over 3 years
  • Has collected 100+ items on Envato Market
  • Sells items exclusively on Envato Market
  • Located in United States
FastNOC says

IMPORTANT

Another search bug/mistake is: if you search for a “scary” it gives you 309 results. But if you want to filter it out by adding two search words “blue scary” it should give only projects which are blue and scary… but now it gives 3101 results….

Every word written in the search acts like a different separate search and stacks results in the same page.

Buyer instead of filtering by adding more tag words actualy makes it harder to find. This is sooooooo wrong.

That is not a bug, it’s correct, and by design. Adding words to a generic search ADDS variables, it doesn’t isolate them.

the same thing will happen in google. Unless you specify to make your search include BOTH of those words.

a typical search will search for any of the words searched. not all of them

EDIT: To clarify.

Search google.

Search criteria: scary “About 53,500,000 results”

Search criteria: blue scary “About 283,000,000 results “

the typical technique to search for BOTH words, in any order, would be to use the term:

“blue AND scary” that’s called a boolean search if you want to search for either of the words, you can also do “blue OR scary”

1617 posts I create things I wish existed
  • Elite Author: Sold more than $75,000 on Envato Market
  • Has sold $75,000+ on Envato Market and is now an Elite Author
  • Won a competition
  • Made it to the Authors' Hall of Fame
+8 more
cyzer says

I remember really well that search did not work this way few months ago. And I think it shoudnt work this way . If you want to find something specific like “fast dark logo” then it wount shorten the list but will increase it instead… what the hell people are you serious :))) its not how the search is supposed to work. The more you type in to search the less and accurate results it should give.

142 posts
  • Has sold $10,000+ on Envato Market
  • Has collected 1+ items on Envato Market
  • Joined us in one of our Envato Live Events
  • Made it to the Authors' Hall of Fame
+3 more
sunchannelmusic says

I remember really well that search did not work this way few months ago. And I think it shoudnt work this way . If you want to find something specific like “fast dark logo” then it wount shorten the list but will increase it instead… what the hell people are you serious :))) its not how the search is supposed to work. The more you type in to search the less and accurate results it should give.

Agreed on this. The more words you have in a search, the more refined the search will be. E.g. If “logo” has 10 items in a search, then entering “fast dark” should find all things “fast dark” within those 10 items.

I think overall the simplified search seems to affect the marketplaces with the largest quantity of items. This is where more advanced search options are more necessary to sort through all the items. VideoHive, AudioJungle, and GraphicRiver have the most number of items. If you check out the respective forums, authors are discussing a sudden drop in sales around the time the new search took place. Only exception to my large item quantity theory is PhotoDen whose forum is not as active.

Anyhow, that’s my humble theory.

189 posts
  • Has been part of the Envato Community for over 3 years
  • Has collected 100+ items on Envato Market
  • Sells items exclusively on Envato Market
  • Located in United States
FastNOC says

I understand these points, but the problem is that’s not the way a normal search works. My example explains it.

the way searching a database works, whatever is put in the query adds to criteria, it doesn’t reduce it. The thought is, searching for ‘blue scary’ means it is going to show you results for anything with either of the words. that is a basic search and that’s the way all search engines work. It’s also the default way Google works. The reason you don’t notice it as much is that Google does an incredible job of ‘understanding’ what you’re really looking for. then again, the complexity of Google’s methods are staggering.

To search the way you’re referring is by phrase, or Boolean which is actually a big change from a normal inclusive search. So if you consider just the load on the db server, doing Boolean searches requires two things. 1. more resources because the query itself is more complicated and 2 it requires additional programming and different capabilities. this kind of things isn’t done by default, as a server administrator you have to enable these functions.

Having said that, it makes sense to do it. This goes back to my entire problem. The searching should be one of the most robust things on the site and demand the most focus. This is the tool that will determine whether a person does or does not find a product to purchase.

I just did a really quick test, and I don’t believe Boolean searches are enabled at all on this site. IO could be wrong, i only did one test. But this is specifically what I’m talking about when i say these are very basic features and should be available even on a very basic data driven site. I’m a little confused why that’s not there already.

Based on the staff response here there are features already in place with built in queries to to semi custom searches by weighting results. not making those features available in switches or check boxes surprises me. But my guess is this is a very basic setup, and very possibly done to keep the taxing of resources down. But that’s purely a guess.

The thing that surprised me was the answer that the servers are in the US. I assumed this was a geo distributed cloud based website but I guess that’s not the case.

Anyway, These things are not luxuries, they are very basic requirements.

2232 posts aspire to create - create to inspire
  • Elite Author: Sold more than $75,000 on Envato Market
  • Has sold $125,000+ on Envato Market
  • Made it to the Authors' Hall of Fame
  • Had an item featured on Envato Market
+8 more
Creattive says

nevertheless, it would be easier to then put all the results that match both search queries in the beginning, as they should be more accurate search results. the way it is now, this doesn’t work. cyzer showed that in his thread.

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

IMPORTANT

Another search bug/mistake is: if you search for a “scary” it gives you 309 results. But if you want to filter it out by adding two search words “blue scary” it should give only projects which are blue and scary… but now it gives 3101 results….

Every word written in the search acts like a different separate search and stacks results in the same page.

Buyer instead of filtering by adding more tag words actualy makes it harder to find. This is sooooooo wrong.

As other have pointed out, not a bug, but it does this by design. For sorting by relevance, this makes a lot of sense. The way the scoring works is that items that have a good match on all words are higher ranked, but items that only match on one word still get included, but further down the list.

However, for sorting by sales/rating etc, you’re right, and it probably doesn’t make as much sense for it to expand the search results, but rather to refine it.

It’s a good point, and something to think about.


I think overall the simplified search seems to affect the marketplaces with the largest quantity of items. This is where more advanced search options are more necessary to sort through all the items. VideoHive, AudioJungle, and GraphicRiver have the most number of items. If you check out the respective forums, authors are discussing a sudden drop in sales around the time the new search took place. Only exception to my large item quantity theory is PhotoDen whose forum is not as active.

Anyhow, that’s my humble theory.

The big challenge with videohive and audiojungle in particular, is that they have multimedia items that are harder to index and search on, and are harder for buyers to scan through to get a feel for what the item is like. With themes, photos etc, it’s often enough to view the preview thumbnail, and scan the description to get a good idea of what it does.

With audio/video, you need to sit through the whole thing to really see if it’s what you’re looking for. So buyers tend to look at fewer items, and buy from the higher ranked items more than on other sites.

What changed a few months back when we launched the new search engine (and what a lot of the discussion was about), was that the old search engine was kind of naive, and didn’t do a good job of fairly surfacing items. So a few authors who had previously “got lucky”, and ended up at the top of the search results for very common terms because that was where the results fell. The new search engine balances that out and does a much better job at surfacing content fairly – but the authors who won the lottery and had top place for very competitive terms lost out. However we get a much better spread of search results that benefits authors and buyers.

We were very careful about rolling that out, and we’ve got metrics to show total sales went up, and conversions of the % of buyers who did a search, and then purchased went up.


To search the way you’re referring is by phrase, or Boolean which is actually a big change from a normal inclusive search. So if you consider just the load on the db server, doing Boolean searches requires two things. 1. more resources because the query itself is more complicated and 2 it requires additional programming and different capabilities. this kind of things isn’t done by default, as a server administrator you have to enable these functions.

Not so much :) The elasticsearch servers can handle that kind of query fine. However elasticsearch gives you a LOT of knobs to tweak to control how search works. The hard part is finding how it should all be set up so results make the most sense.


I just did a really quick test, and I don’t believe Boolean searches are enabled at all on this site. IO could be wrong, i only did one test. But this is specifically what I’m talking about when i say these are very basic features and should be available even on a very basic data driven site. I’m a little confused why that’s not there already.

Technically it’s using a dis max query. The search term is broken into tokens, and a subquery is done on each field (name, description, tags etc). The subqueries then get merged together and sorted depending on which subquery generated the highest score.


Based on the staff response here there are features already in place with built in queries to to semi custom searches by weighting results. not making those features available in switches or check boxes surprises me. But my guess is this is a very basic setup, and very possibly done to keep the taxing of resources down. But that’s purely a guess.

We’d love to expose it all. We tweak it a lot with the back end queries that are run. The tricky part is figuring out (and building) how it is presented and controlled by users, and how it gets passed to the backend. Not hugely challenging problems themselves, but there are lots of other things we need to build too, and work can only get done as fast as it gets done :)


The thing that surprised me was the answer that the servers are in the US. I assumed this was a geo distributed cloud based website but I guess that’s not the case.

The app servers/database/search servers are hosted in a datacentre in the US. That’s where all the dynamic content gets served from. We haven’t had a need to fully distribute the dynamic infrastructure (yet).

Static assets (.js, .css files, images etc), and large files such as item uploads/downloads etc are cloud based on Amazon S3.

The servers aren’t in Australia as some might have expected (even though we’re an Australian company). The infrastructure in Australia is generally less competitive and more expensive than the US. Plus it’s a huge latency cost to hit an Australian server when most of the website users are in other parts of the world. That big “Pacific Ocean” thing is a pain for low latency networking ;)

189 posts
  • Has been part of the Envato Community for over 3 years
  • Has collected 100+ items on Envato Market
  • Sells items exclusively on Envato Market
  • Located in United States
FastNOC says


IMPORTANT

Another search bug/mistake is: if you search for a “scary” it gives you 309 results. But if you want to filter it out by adding two search words “blue scary” it should give only projects which are blue and scary… but now it gives 3101 results….

Every word written in the search acts like a different separate search and stacks results in the same page.

Buyer instead of filtering by adding more tag words actualy makes it harder to find. This is sooooooo wrong.

As other have pointed out, not a bug, but it does this by design. For sorting by relevance, this makes a lot of sense. The way the scoring works is that items that have a good match on all words are higher ranked, but items that only match on one word still get included, but further down the list.

However, for sorting by sales/rating etc, you’re right, and it probably doesn’t make as much sense for it to expand the search results, but rather to refine it.

It’s a good point, and something to think about.


I think overall the simplified search seems to affect the marketplaces with the largest quantity of items. This is where more advanced search options are more necessary to sort through all the items. VideoHive, AudioJungle, and GraphicRiver have the most number of items. If you check out the respective forums, authors are discussing a sudden drop in sales around the time the new search took place. Only exception to my large item quantity theory is PhotoDen whose forum is not as active.

Anyhow, that’s my humble theory.

The big challenge with videohive and audiojungle in particular, is that they have multimedia items that are harder to index and search on, and are harder for buyers to scan through to get a feel for what the item is like. With themes, photos etc, it’s often enough to view the preview thumbnail, and scan the description to get a good idea of what it does.

With audio/video, you need to sit through the whole thing to really see if it’s what you’re looking for. So buyers tend to look at fewer items, and buy from the higher ranked items more than on other sites.

What changed a few months back when we launched the new search engine (and what a lot of the discussion was about), was that the old search engine was kind of naive, and didn’t do a good job of fairly surfacing items. So a few authors who had previously “got lucky”, and ended up at the top of the search results for very common terms because that was where the results fell. The new search engine balances that out and does a much better job at surfacing content fairly – but the authors who won the lottery and had top place for very competitive terms lost out. However we get a much better spread of search results that benefits authors and buyers.

We were very careful about rolling that out, and we’ve got metrics to show total sales went up, and conversions of the % of buyers who did a search, and then purchased went up.


To search the way you’re referring is by phrase, or Boolean which is actually a big change from a normal inclusive search. So if you consider just the load on the db server, doing Boolean searches requires two things. 1. more resources because the query itself is more complicated and 2 it requires additional programming and different capabilities. this kind of things isn’t done by default, as a server administrator you have to enable these functions.

Not so much :) The elasticsearch servers can handle that kind of query fine. However elasticsearch gives you a LOT of knobs to tweak to control how search works. The hard part is finding how it should all be set up so results make the most sense.


I just did a really quick test, and I don’t believe Boolean searches are enabled at all on this site. IO could be wrong, i only did one test. But this is specifically what I’m talking about when i say these are very basic features and should be available even on a very basic data driven site. I’m a little confused why that’s not there already.

Technically it’s using a dis max query. The search term is broken into tokens, and a subquery is done on each field (name, description, tags etc). The subqueries then get merged together and sorted depending on which subquery generated the highest score.


Based on the staff response here there are features already in place with built in queries to to semi custom searches by weighting results. not making those features available in switches or check boxes surprises me. But my guess is this is a very basic setup, and very possibly done to keep the taxing of resources down. But that’s purely a guess.

We’d love to expose it all. We tweak it a lot with the back end queries that are run. The tricky part is figuring out (and building) how it is presented and controlled by users, and how it gets passed to the backend. Not hugely challenging problems themselves, but there are lots of other things we need to build too, and work can only get done as fast as it gets done :)


The thing that surprised me was the answer that the servers are in the US. I assumed this was a geo distributed cloud based website but I guess that’s not the case.

The app servers/database/search servers are hosted in a datacentre in the US. That’s where all the dynamic content gets served from. We haven’t had a need to fully distribute the dynamic infrastructure (yet).

Static assets (.js, .css files, images etc), and large files such as item uploads/downloads etc are cloud based on Amazon S3.

The servers aren’t in Australia as some might have expected (even though we’re an Australian company). The infrastructure in Australia is generally less competitive and more expensive than the US. Plus it’s a huge latency cost to hit an Australian server when most of the website users are in other parts of the world. That big “Pacific Ocean” thing is a pain for low latency networking ;)

Thanks for that response!

yeah I have to just guess on what you’re running, I mean obviously I don’t know what’s happening under the hood, and frankly I would have been veyr surprised if this was just running vanilla stuff. So that doesn’t surprise me.

Dis max makes a LOT of sense. I guess the reality is you’ve just not gotten to the point of exposing the possibilities to us yet, but i have to say over and over, it is costing me a TON of time (and when i think that way it boils down to dollars) having to hunt and peck for things, and iit’s causing sales loss. There’ve been multiple times I’ll be weeding through projects and see one that could have worked for me. The problem i have is the same as it with theme forest. Well sort of. I could task our team to write themes that mirror the functionality of the incredible wordpress themes. And even though it frequently looks like all of the themes are built on the same core template i know they aren’t but when we came into the retail side of things it severely restricted our capability to design the themes i see on TF. now, while I’m new to the After Effects crowd, I’ve taken courses and know the fundamentals and cxan usually get what i want, but the bottom line on both is it would cost me many thousands of dollars to have my teams create a wordpress theme that could compete with the functionality we find here. So it leaves us no choice, retail clients aren’t going to pay our bill rate to put a theme together. I’d venture to say they wouldn’t pay even 25% of it. So we are at the mercy of the community here. That’s why i’m so adamant about this topic because while I know it’s costing Envato sales, I can directly see the lost work it causes me not being able to find things.

by the way, you asked about how i search on AJ for by length? If I’m looking for a track that’s say 1:30? i’ll search with this: browse/attributes/length/1:30 but i don’t believe that attr is being used on VH

regarding servers, I just assumed you’d be using S3. i figured since most of the authors seem to be outside the US (based off nothing more than the accents i hear in videos included in projects, so it’s certainly not an educated guess) you’d get more traffic outside than inside.

Well, that and I’m frankly in love with S3. :) We have servers in many different locations, but our business has changed so much since grid started then the cloud. We’ve still got racks co-located in a few datacenters, but have been migrating to distributed.

Side note, I built an entire render farm on S3 so i could upload projects to render there rather than doing it on my PC. since they now offer graphics instances (although i’ll admit, i didn’t utilize the gpu stuff because of cost, I just uploaded projecst that didn’t need tons of cuda cores, which pretty much means anything that isn’t using element 3d.. most projects dont use gpu to render, so base windows servers work fine. incredible how big you can make a farm, and how fast. the built buckets and pipelines for transcoding. was pretty slick. But i had to shut it down because one of our biggest clients (ironically a security agency) had one of the guys get hacked on S3 and the hackers built HUGE farms to mine bitcoins and the cost was incredible by the time it was spotted. yikes!

35 posts
  • Has referred 1+ members
  • Sells items exclusively on Envato Market
marketopia says

Another problem, that is apparently not going to be fixed by the upgrades, is sorting by sales. I don’t know if you realize this, but sorting by sales is virtually pointless. Here’s why. If I search on themeforest for a WordPress theme, and I sort by sales, I’m going to find that the results don’t show me what I need. It shows me the overall sales. Well that’s not relevant. Most times you’ll end up getting results from themes that were built YEARS ago. Because they’ve accrued a higher number of sales than anything new. But what if I’m looking for something that’s really hot right now. I want to search and sort by the sales, but I don’t care about results from projects that are over a year old, because they are much less likely to have the updates of the newer technology. This is critical in this day because with HTML5 things are very different. Qualifiers should be much broader and I for one really need to be able to chain variables. I want to sort by sales in the last 6 months. Or maybe I want to sort by sales in the last 6 months, but only projects that have woocommerce. That’s not possible. And I don’t see any of that changing with these features. The sales and ratings sorting are pretty much useless because they cannot be qualified. Sort by ratings and you end up with pages of results from files that are year’s old, sales, same result.

If you use CodeCanyon or ThemeForest, you can consider using the advanced search and filtering system on Marketopia.net :)

by
by
by
by
by
by