Closed Bug 765768 Opened 12 years ago Closed 12 years ago

[research] analyze elastic search unified search results

Categories

(support.mozilla.org :: Search, defect, P1)

defect

Tracking

(Not tracked)

RESOLVED FIXED
2012.12

People

(Reporter: willkg, Assigned: Matt_G)

References

Details

(Whiteboard: u=dev c=search p=2)

Once elastic search unified search results lands, we need to:

1. analyze the search results compared to the elastic search bucketed search results

2. come to a concensus of: a) they look good enough, or b) they need more work

3. in the case of b) we need to write up new bugs to implement necessary changes to make the search results better


The outcome of this bug is analysis from number 1 and either thumbs-up or more bugs.

Putting this in the 2012.12 sprint as a P1 and with 2 points, though that's kind of a guess and this is not really an implementation bug but rather an analysis-and-talk-with-other-people bug. Also, it might make sense for Matt to do the analysis.
cc:ing Matt on this since I'm pretty sure he's going to be driving the work here.
Assignee: willkg → nobody
Fixing the summary so it doesn't look like the other "elastic search unified search results" bugs.
Summary: [research] elastic search unified search results → [research] analyze elastic search unified search results
Assigning this one to Matt.

Matt: If you need help or anything, let me know.
Assignee: nobody → mgrimes
Fantastic Will. I am meeting with Michael Verdi later today. We'll put together a testing plan and get moving.
Hey Will. We finished our initial testing. I am still hoping you can explain what happened in that example search. That issue aside, we were able to identify a few issues. I'll open new tickets for them.
Depends on: 767540
Depends on: 768240
Depends on: 768242
Just to clarify, you're adding blockers to this bug which I infer to mean that this bug requires those things to be complete before we can close this. And that means that we're staying with bucketed search until all of these things are implemented. Is that correct?

The reason I ask is that unified search is almost assuredly better than bucketed search already. Like I explained in our discussions on Vidyo, I was hoping we could do some minor tweaking with the existing infrastructure, then switch to unified search for everything and ditch all the bucketed search code. Then we could continue honing unified search going forward.

If it's really the case that the unified search results are bad enough that we need to do things we can't currently do (like bug #768240), then so be it. But it means we're not switching to unified search results for a while.

Also, has any of the analysis done for this bug so far been documented anywhere?
Hey Will. I think that with the removal of contributor posts it will be as good as bucketed search or better. We are documenting our testing in a spreadsheet right now and we'll take another run at it later this week to confirm. I was adding those other bugs to get it really crisp and accurate based on what we've seen so far in our testing. I don't think they'll be blockers before pushing it live. They can probably come further down the road, but I'll need confirmation from a few other folks before I can make that call.
I'd suggest that we push this to 50% of our users as soon as possible, take a look at the CTR for both and then make a decision to move to unified search or not. What do you guys think?
That sounds good to me Kadir. I'll continue to test on my end as well to look for future improvements and content tuning opportunities.
Given that, I'm going to nix the uncompleted bugs from the blocker list for this bug, change the waffle to 50% and then mark this as FIXED.
Status: NEW → RESOLVED
Closed: 12 years ago
No longer depends on: 768240, 768242
Resolution: --- → FIXED
You need to log in before you can comment on or make changes to this bug.