bloodhound-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Joe Dreimann <>
Subject Re: Query Builder (Search / Custom Query)
Date Wed, 20 Mar 2013 23:16:29 GMT

On 20 Mar 2013, at 20:53, Olemis Lang <> wrote:

> On 3/20/13, Joachim Dreimann <> wrote:
> [...]
>> On 14 March 2013 19:47, Andrej Golcov <> wrote:
> [...]
>>> We can set default product filter based on active product and use
>>> global search if user navigates directly to global search url - I
>>> think, Olemis suggested something like this.
>> I believe search should always be global unless the user chooses otherwise.
> what does this mean exactly ? Let's focus on the following situation .
> User reads let's say a wiki page in product P and searches for word W
> by using search input box in the header . In your opinion is there any
> value for getting results from other projects , even potentially not
> provide any result for the active project ?

Yes. Given your scenario:

- If the user was reading the page it may have reminded her of W in another product (or user
is uncertain about the product it's in), not retrieving that information = bad (initially
the user may even think it doesn't exist at all)
- The user may have just had the page open from reading earlier, but coming back to the tab
now after an hour she may be looking for something unrelated = bad
- The same action (search) having changing scope without an explicit user action seems bad

The first two are common use cases for me and ones I often see others do. I may suffer from
confirmation bias here though so do let me know if that's not your experience.

There's a stronger argument for me though, and I'd love to see log data to prove or disprove
it. I believe most searches are navigational, essentially to named objects that users have
seen or interacted with in the past, or that they know about by other means. Users tend to
find these through query negotiation by refining their query to narrow it down, but crucially
rarely to broaden it unless they have high certainty that it exists and high determination
to get there. We need to design for people for whom that's not the case, that means full scope
by default to me.

Don't mistake that for me being certain, I would much prefer to test both approaches with
users. If you think that's feasible I may be able to set up such a test.


> That's more than confusing afaict ; unless the user explicitly checks
> a selector to search anywhere ... that's another subject ;)
>> Our ranking should ensure that most relevant items move towards the top,
>> and a data point for this could be recently viewed products.
> I agree with this when it comes to explicit global search view .
> [...]
> -- 
> Regards,
> Olemis.

View raw message