Is there a more comprehensive understanding of the way confluence ranks search results?

[Jake Lown]

I have read https://confluence.atlassian.com/display/CONF42/Ranking+of+Search+Results but it seems as though I am not getting the expected results in ranked searches. Is there a more comprehensive and definite way in which one can influence the way search results appear? Is there a way to either break down the existing ranking system further or is there a way to influence the way search ranks appear?

[DebbieR · Accepted solution]

Hi Jake,

I was given this info from Atlassian when I asked a similar question a few months ago. Hope it helps.

Here's a very brief run down on how search results are weighted. There are no parameters to manage the ordering of the search results. 
 

DOCUMENT AGE

Newer Documents will be given a slightly higher score than older documents. Keep in mind that doesn't mean that newer documents will appear first, just that they give a slightly higher weight.
 

TITLE, LENGTH OF TITLE, FREQUENCY OF SEARCH TERM

Confluence will look at the number of words within a title (excluding minor words, such as "the", "and" etc), and will check how many times a word appears in the title. Titles that are shorter (around 4 words or less) will rank higher than titles with more than 5 words.
 

CONTENT

As with the title, Confluence will look at the number of words within the content, and check how many times the number of words appear in the content. Longer pieces of content should rank lower than shorter documents with the same number of references.
 

POPULARITY

Pages with the same content will be ranked according to their popularity within Confluence. A page that has more incoming links will be ranked higher than other pages that share similar content. 

Icon

The search results were documented previously by Atlassian (refer to: Ranking of Search Results and to Confluence 5.2 Release+Notes for further information), but due to changes to search in v5.2 it was decided not to document the search - as it's quite complex).

[Jake Lown]

This was perfect Debbie, thanks so much

[Alexa Tolan]

how is this possible that you don't document the Search because it's too complex? nothing should stop you from documenting such a crucial feature like the Search tool. My colleagues give me negative feedback about the Search tool all the time... How could I help them find their way if you don't provide me with the complete knowledge after each release?

[Nic Brough -Adaptavist-]

I'd tend to agree with that, a lot of people complain about Confluence search not returning things in the right order for them, and without understanding what the search is doing, it's hard to explain to your users. "We don't know because Atlassian don't tell us" is not a good explanation. Although I'd rather have effort spend on getting the search to actually work than document why it doesn't work well.

[Alexa Tolan]

Nic, it's true that Atlassian should invest time in improving the search tool, but to accomplish this they need the user's feedback. at least this is how I see things. but then how can we (the users) give them consistent feedback if we don't undestand in the first place how the search engine was designed to work?

[Nic Brough -Adaptavist-]

You don't need the full intricate details of what it's doing to be able to explain that it doesn't work. I'll fall back on my usual basic analogy - a car. Your car has probably been sold to you with a pile of claims and statistics about pollution emissions, fuel consumption, noise levels and so-on. You don't need any understanding of the workings of an internal combustion engine to be able to explain to the manufacturer that you're getting poor fuel consumption, black smoke from the exhaust, a funny rattle from the engine when it's going over 30 etc. On Confluence search, it's mostly poor for me because it doesn't have any useful way to promote "canon" pages. I had a client for a long time who had an unusual login system. Because it was unusual, people often took notes on it for their personal use, in their personal space. Teams documented it for new starters in their team spaces. All a complete waste of time, because there was a small set of *really* well written documents that covered it perfectly. But the pages were always buried because Confluence prioritised "recent edits" over "stuff that's actually useful". In this case, it's a recursive problem - because the search was broken, people never found the right pages, and because they didn't find them, they wrote their own, which made it worse because the more recent (often poor) duplicate pages rose to the top of the list. I don't need to know how search works internally in order to point out that search is broken in that case.