Popular files are already advertised well on the marketplaces by their nature , naturally.
We should be more protective for the Non-Popular files.
the first item ( the one that is not a video player) have “video” as tag and also “video” in the description and, is also placed in Images and Media category and also a lot of sales
Yeah that listing is easier to understand than the 6 items averaging 13 sales and no rating. But still, this is why search probably needs some hands-on manipulation—a staff member dedicated to manually improving search. Every day they work off a list of the top search terms, evaluate the results and make adjustments. In the example I posted, the lightboxes and galleries don’t necessarily need to be removed, but they need to rank lower than actual video players.
CodingJack saidYep that makes sense and would be fair ( and also a improve the users search experience )
dtbaker saidAssuming this is an extension of the user’s original search, logically it would be the next 7 search items. So if the user searches “ice cream”, and clicks on your item which is ranked #20 for “ice cream”, the 7 thumbnails would be results #21-27 from the original search listings. Hopefully this is how it works. Because that truly adds value to the user’s search experience, and is also an unbiased method that I think we’d all be satisfied with.
It will be interesting to see how competing items are sorted. Most sales, ratings, newest, popular, – I don’t think there is a perfect way (maybe the visitor can have options to sort the competing results for a better experience? eg: what items are similar to this but with the highest rating)
+1Plus, it should be only displayed if a customer visits item page through search, and not by other means.
Great, now we’re getting somewhere Authors are never against improving the user experience. We just don’t like back-door ways of padding other people’s pockets
doru saidWhat if everyone wants to check both?
An opt out feature is an interesting approach
this should work like this
checkbox a: “Disable other authors items on your item page”checkbox b: “Force your items on other authors item page”
then we will be like now.
it may look like a joke but think about it. If there are authors who believe this change is ok then they will have no problem in leaving the “checkbox a” unchecked
Thanks for that! its just getting better and better here!
Any updates from the Staff regarding our feedback on ‘related items on Item’s page’ ?
It’s unbelievable something is going to change in that matter after years. I hope it will improve.
Filtering by “last updated”If you don’t want to start an updating war, and triple length of queue – please don’t implement this.
FAQ / Comments searchJunction of those two is a must. With higher priority in search results for FAQ section.
Similar items on a product pageI’ve noticed most of commenting authors focus on the users will go away from my page aspect, and forget about users will come to my page aspect. As long as system will not favor any item, and would suggest them equally – it means competing with quality of your product + item page – I’m in.
Although, opt-out per item is a good idea for those who bring traffic with external campaigns and don’t want to share profits of it.
Anyway, please focus on search results quality in a first place.
Thanks again everyone for your input.
Just letting you know we are evaluating a few options and haven’t made any final decisions yet, will depend on technical feasibility and best approach for the community. This was pretty much a free feature for us which took very little effort to implement and we had good reason from examining user behaviour that this would be beneficial to both authors and buyers. That said we don’t want to prioritise making any changes to it that will divert our attention from the more important work – like improving relevancy (currently working on but will be ongoing) and implementing comments search (currently working on).
We have a long list of improvements we want to make, I understand they are long overdue so we are working as fast as we can . Thanks everyone for your top 5!
I thought I’d jump in and provide a bit of technical background on the search work as a marketplace senior developer.
The new team
Search has had continual work done on it, so it’s not like we let it die, and then suddenly picked it up again.
For a long time there was 1 small development team working on ALL the marketplace code. As the sites grew, it became one big team, then a bigger team, then bigger still. At a certain point, we realised this wasn’t scaling, and we needed to focus better. So we’ve split into smaller teams focused on different areas. Search being one of them.
What this means for search, is that instead of development work going into “one big bucket” to be picked up by various developers, we’ve now got a smaller group of us focusing JUST on search. We can spend more time thinking about it, and doing it properly.
The new platform
We want to build more cool and useful features based on what the community wants and needs. However, we’ve got a lot of ugly legacy code and platforms to deal with. It has gotten to the point where we had 3 different code bases for different parts of search on different parts of the site. All based by 3 different search engines. This makes working on it slow and error prone.
In addition to the technical debt, we were also hitting scalability problems. SOLR powers most of the search, and it is struggling under the load.
I know personally. I’ve been on call when it dies 4 times a week at 3am, and needs a kick to restart it and get it going again.
In addition, response times from it have steadily been getting worse as we get more data, and more traffic.
(I’m not saying SOLR is a bad solution – it’s still pretty good, just our implementation of it was buckling under the strain, and needed some fixing)
So a few months back we started a consolidation project to clean all this up. We’ve moved from a frankenstein mix of Sphinx, SOLR, and elasticsearch to just a clean rebuild using only elasticsearch. The vision for this project is: “make it EXACTLY the same as it was before, only cleaner, and easier to work with behind the scenes”.
So while you don’t see any new features initially, it makes it MUCH MUCH easier to build them from now on. Item comments are the top priority. Scoping out this work is probably an order of magnitude easier to build than it was before the consolidation work.
We’ve also engineered it to scale much better. The philosophy the overall marketplace dev team to scaling is “10x” – design everything to work at 10x the data and traffic volume. elasticsearch has some amazing ways of doing this which we’re taking advantage of. Check out this interview with the github elasticsearch team for a bit more background on what you can do with it. We’re not at that scale, but we’re using some of the same techniques.
More like this
There are a whole lot of issues already discussed, which I’m not going to go into.
However, I will add that part of the decision to turn this on, is that we basically get it built in for free with elasticsearch. We had already played around with it on Photodune and Audiojungle, so the technical time spent implementing it was very low (like “10 minutes deleting an
if block in a rails view” low).
So technically, it’s a no brainer to enable. BUT there are obvious a lot of bigger issues which need to be considered, and these are still being discussed.
MSFX saidOnly in ActiveDen, no one can sense the damage. A great place for nuke test!
yeah, starting with Activeden rocks, it’s where the party started baby
Actually, that’s pretty much correct
Activeden is small, gets low traffic, so it’s a good place to turn it on initially. But roll out to other sites will be quick. We’re probably talking days here.
We’ve had it dark launched and running with production traffic for a month or so though. We log all the search queries, and have been replaying them against the new search engine offline.
So we’re already pretty confident that it handles real world load, and that it handles all the crazy weird things that real humans (and automated bots scraping us) do.