Search algorithm: proposal to improve

Hi.
I know that many of you have thought about how search works, and how to appear on the first results page by default (the Best Match).

My observations lead me to this vision of the situation:
1. Among the authors there is a strong opinion about what the most popular searches are words like “inspiring”, “motivation”, “happy”, “upbeat”, “trailer”, “epic” etc.
2. Aspiration to hold out as long as possible in the first positions of search results for such requests motivates authors to call their tracks using only these words, or even a single word.
3. This leads to the appearance of dozens of tracks having the same or almost the same name. As a result - this tactic stops to ensure a long position tracks on the first pages of search results.
4. A set of high-quality and promising tracks go unnoticed because they are at the end of the search results or do not get into it at all.

As I propose to solve this problem:
1. The default sorting of search results to make the display of tracks in random order. This algorithm takes into account the availability of the requested word is not in name only, but in the tags. At the same time the results are displayed exclusively in a random order.
2. The sequence of the search results are constantly changing (more often is better). In my opinion it has leveled the ability of authors to be heard by buyers and it will be fair.

What do you think about this?

8 Likes

I agree with this. At this point, at this enormous rate of new incoming tracks, the only solution is to use some kind of randomization in the ordering of the relevant search results. It doesn’t have to be all the results, but maybe 1/3 or close to that.

Tags (and ridiculous descriptions) are just as overused as the titles so going after tags more or less wouldn’t change anything.

1 Like

There are TONS of great tracks that i just keep listening and wondering how small amount of sales they got…And when i read the name of the track, it tells me everything…Buyers get used to type the words you have mentioned into the search bar and they only get “Upbeat, Uplifting, Ambient, Inspiring, Motivation” tracks etc…But what about those with more creative name and even more creative music? I felt this on my skin because whenever i name the track that is not “commercial” name here, i got no sales or just few…And thats the destiny of ALOT tracks here…I think that Envato should seriously think about this proposal and suggestion and make something out of this…Alot of good authors are hiden because of this…I had alot of ideas how to resolve “special name” authors but i will create a new thread for it…Let’s see what developers will say about this.

4 Likes

Hundred percent behind this idea!

I recently asked a question about the search for customer support. I was answered that they work on it. I described them to my proposal that you have read the above and asked what they think about it. I didn’t receive the answer.

Some of my colleagues tend to think that this situation is beneficial for Envato. They say something like this: Why Envato make efforts and change something in relation to drawing attention to the tracks having a few sales. Tracks from the first pages of search created by authors favourites so make good profit.

I do not know how true this is. But anyway such situation won’t lead to positive results in the long term. If this is true, whether that is simpler to cease to accept tracks of those genres which there is a lot of?

When I spend 2-3 days to create the track, I hope that it will buy at least 3-4 times. Otherwise, the profit that it brings does not justify the time spent. I think many will agree with me. The current default search algorithm poorly motivated to download new tracks certain genres.

I started this thread in the hopes to draw attention to this situation and to change something. Thank you for your comments, let’s try to do it together!

1 Like

Today I received a reply from customer support. They agree that the current situation is not optimal and are looking for ways to improve.
Hopefully that will change as soon as possible, good luck!

1 Like

I guess what you really want from a search algorithm is to link buyers up with tracks they might want to buy. As an experiment I tried looking for various types of tracks as if I were a buyer and found that the results were often highly random and not terribly helpful. Obviously if you are looking for the standard corporate/motivational style you will be spoilt for choice but if you’ve got anything more specific in mind I’m not sure the search engine as it currently operates always helps you find appropriate tracks.

I’d be very interested to hear more about how things seem from the point of view of the people actually buying the music - after all the primary point of the search engine is to help them!

1 Like

+1 for the suggestions and it’s what @PaulGraves says: buyer perspective.
What if more specific searches ar shown first?
So if you search for “positive”, “upbeat” and “cinematic” (f.i.) together, the track with the most tags, title, item description and categorie is shown first. Similar results: random order to be fair.
I’m not a seo expert :-), so i’m sure bether algorithm’s are possible.

If buyers want a credible playlist to be put in front of them the track titles MUST be removed from the search engine algorithm. Buyers should be instructed to type 3 to 7 KEYWORDS that describe

  1. The main genre they need - corporate, rock, indie, funk, cinematic, etc
  2. Then descriptive adjectives: happy, sad, upbeat, epic, energetic etc.
  3. Then more “environment” style words like beach, summer, night life, bar, outdoors,
  4. Then even tempo…slow, medium, fast

I think customers need to be told how to get the best search results

I just tested this search from 5 keywords and I have to say the result displayed were pretty good:

rock, energetic, fast, cool, sports

then I tested

inspire, background, video, minimal, pop

Strong encouragement of buyers to use at least 5 keywords (or just mandate it) may be the simple fix. I highly doubt buyers use 5 keywords.

1 Like

Wow, I just searched these keywords to test my own theory for a track:

cinematic, waltz, strings, beauty, pretty, nature

Then got

“You’ve ventured too far out into the desert! Time to head back.
We couldn’t find any results for your search.
Use more generic words or double check your spelling.”

So I eliminated the word “Nature” and a 1 track playlist was displayed. I guess the search function DOES need a lot of work. These were words I was going to use to describe and upload a new track. I guess I’ll forget about uploading that track.

2 Likes

I used to have exactly the item which would fall under that description, by the way. But I didn’t get even one sale on it and decided to take it down XD

By what logic? That just means that no one has used all of those tags/words in their entire item page. Why on earth would you not upload a track based on that?

1 Like

Hi sixthulip!

The problem with randomising all of the AJ catalogue for a specific keyword search is that many really old tracks with no sales will clog up the first result page. The good thing about search right now is that it takes upload date and sales into account, so most of the page 1 results are actually either fresh or selling. Consider this, search for “inspiring” and you get 45,000 results. If you randomise these results, the chance of your hot new track being on the front page is now 50 / 45,000 ≈ 0.1 %. Even after 1,000 searches your brand new music will only show up as one of these 50 tracks ONCE. This is not better for anyone in the long run, there simply is not enough sales to make every track “pay off” for every author. Also, buyers would not likely want to have to listen to year-old (or 5-year-old) tracks by default.

An alternative I’ve suggested many times is to alternate search results on every other line, like result 1 is primarily title weighed, 2 is tag-weighed, 3 is title weighed, 4 is tag weighed and so on. That will at least take some focus off the titling and bring more interesting titles into search.

Other than this, I’d really like AJ to consider offering curated music supervising as I recently mentioned here :sunglasses:

1 Like

"…many really old tracks with no sales will clog up the first result page."
The fact that the track is old does not necessarily mean that it’s bad, is not it? In my opinion, now the situation is such that the track becomes “old” if it was not purchased in the first day or two (for the popular category).

"Also, buyers would not likely want to have to listen to year-old (or 5-year-old) tracks by default."
For this there is a filter “New items”

"…1 is primarily title weighed"
I think now is not many authors have come up with original names. They are faced with a choice: to call his track “Inspiration” and stay on the first page of search a few days or to come up with an original name and not to fall in the search results at all.

1 Like

I uploaded a new Corporate Motivational track that I decided to call “Lifting” to be descriptive and close to the popular search word “uplifting”. I just did a search experiment that I think has disturbing results. I tried a couple of search variations such as i) typing 3 (out of 3) identical tags including identical name, ii)similar name “Uplifting” plus 2 identical tags, etc. Most of these were reasonable in placement of my track. The search that bothers me is this: I searched on 3 words: Lifting, wholesome, and positive which include identical name (lifting), one identical tag (wholesome) and one tag that I did not include (positive). For this I got “You’ve ventured too far” and ZERO results.
Evidently the search return is set for 100% match of the tags even when I have the exact name in the search! Also this says there is no graceful decrementing for less than 100% match, even when there are zero returns! I had a lot of other tags that are positive-oriented (e.g., uplifting, inspiring, confident, hopeful, optimistic). How many searches are we losing because we failed to place one more tag word in our upload? Thoughts/comments are appreciated. Am I not thinking about this correctly?

1 Like

To me this is completely normal. So now you want the search engine to return results for tags that are NOT included? Oh my, the mess…

1 Like

Thanks for the response prestashopthemes, but that’s not necessarily what I’m saying. I think there should be some search value (albeit down-graded) in the example I posted; perhaps down-grade it by 33% because 1 out of 3 tags were missing.

Note that my biggest issue with my posted example is that the searched title was exact AND I had 2 out of 3 tags exact in the search. Yet the search results did not find my track at all - or any other for that matter. Surely in this example there could be a second tier of search results that compete with each other based on title match and 2 correct tag matches.

In a 315K audio library, why would AJ want to return a search null set to a buyer for the case when one out of several tags he/she inserted is not matched?

3 Likes

I guess my point is that if Envato can’t deliver a playlist based on the 5 or 6 words: cinematic, waltz, strings, beauty, pretty, nature that describe my track,why should I try to sell it on this market?
The track has a cinematic quality, it’s in 3/4 time, features strings, sounds pretty and displays "imagery"of beauty and I think it would be well suited for a nature visual. Prestahopthemes, what would you title the piece and what keywords would you chose? You seem to know a lot about search engine algorithms.

I fail to understand your logic. If you upload your track using those nice keywords and then search for them, Envato will deliver a list of one item: your new track.

I still am amazed that fairly common keywords: cinematic, waltz, strings, beauty, pretty, nature
Do not produce a playlist. People use these keywords all the time. I just tested these 6 again today and a playlist is not produced. Why would that happen? Does this mean no one uses all 6 of these keywords for a given track: cinematic, waltz, strings, beauty, pretty, nature? Even if that is the case…why isn’t the algorithm set up to deliver a playlist based on 3 or 4 of these 6 words? When I eliminate nature and pretty, I did get a very relevant playlist.

The entire point of this discussion is to offer ways to:
Eliminate Track Title entirely from search and therefore promote more interesting titles to look at for buyers
Prevent 75,000 Inspires, the inspire, inspiration, motivations, upbeats, and happy’s and upliftings from hitting the market.
Improve search results for customers who seem to be dissatisfied.
Teach customers how to perform a search, they need their hands to be held. What genre do you need? Describe the feeling of the music you want? How fast or slow do you want the music to be? What type of project will the music be used in?

If the buyer has to systematically answer all of these questions every time they seek a new track, they will have a much improved shopping experience and get tracks they need faster. As it is now, the market is still mostly based on “just go to popular files and buy the tracks titled inspire, inspired, inspiration, corporate, motivation ,upbeat, epic and happy.” If a change does not happen, Every author in audio jungle will title their tracks only those words.