Closed Bug 1835267 Opened 3 years ago Closed 3 years ago

Back/Forward page actions fail to record abandoments if the targets are followed-on SERPs

Categories

(Firefox :: Search, defect, P2)

Desktop
All
defect

Tracking

()

RESOLVED DUPLICATE of bug 1836507
Tracking Status
firefox-esr102 --- unaffected
firefox113 --- unaffected
firefox114 --- unaffected
firefox115 --- affected

People

(Reporter: aflorinescu, Unassigned)

References

(Blocks 1 open bug)

Details

Found in

  • 115.0a1 (2023-05-25)

Affected versions

  • 115.0a1 (2023-05-25)

Tested platforms

  • Affected platforms:
    - Windows 10
    - Mac 13.2.1
    - Ubuntu 22.04

Preconditions

  • Have VPN connected to CA/US.
  • Google as default search engine.
  • browser.search.region set to CA/US
  • browser.search.serpEventTelemetry.enabled set to true.
  • browser.search.log set to true.

Steps to reproduce

  1. Open Firefox
  2. Open a new tab, focus address bar and submit shoes as a search query, submitting the search.
  3. From the incontent search box, refine the search adding 'for', submitting the search.
  4. From the incontent search box, further refine the search adding 'men', submitting the search.
  5. Press the browser's Back button.
  6. Press the browser's Forward button.

Expected result
5. Abandonment event logged
6. Abandonment event logged.

Actual result
5. No abandonment event logged.
6. No abandonment event logged.

Regression range

  • N/A

Notes

  • This applies for both Google/Bing SERPS, although on Bing special attention needs to be paid to bug 1528958
  • If enough back and forward actions are done on the follow-on SERPS, the abandonment will start to work - again, not particularly sure if that's the case on BING due to bug 1528958
  • this bug doesn't apply for persistent search refinement

Just to confirm Adrian, the test does two refined searches within the in-content searchbox, and then after waiting for the page to fully load and for ad components to be recorded, it then presses the back button, waits for the page to load again and for ad components to be recorded, and then goes forward? I guess I just want to make sure we're waiting for the components to categorized because we currently don't record an abandonment for pages that didn't finish categorizing the page for ad components.

(In reply to James Teow [:jteow] from comment #1)

Just to confirm Adrian, the test does two refined searches within the in-content searchbox, and then after waiting for the page to fully load and for ad components to be recorded, it then presses the back button, waits for the page to load again and for ad components to be recorded, and then goes forward? I guess I just want to make sure we're waiting for the components to categorized because we currently don't record an abandonment for pages that didn't finish categorizing the page for ad components.

I guess it makes sense, pretty sure the page was not fully loaded for the back/forward actions. @cbaica, can you confirm this, seems that this is an invalid bug then.

Flags: needinfo?(cbaica)

I think we can keep the bug open as a constraint of the feature as it might be something we want to improve upon.

I've investigated the issue a bit further and my findings are a bit weirder.

On step 5, no abadonment event is logged, but if I press back again an abandonment event logged. Nothing on forward operations though.
I'll attach a recording uploaded to a gdrive, so it's a bit higher resolution so you can check the console log along with the actions performed.

Flags: needinfo?(cbaica)
See Also: → 1836507

Ah thanks for recording that. It looks like we're sometimes counting forward/back movements from one SERP to another as a non ad click.

This is due to a change we landed a few days ago. I've added a new bug related to it: https://bugzilla.mozilla.org/show_bug.cgi?id=1836507

Priority: -- → P2

Hi Adrian, can we try re-testing this? We landed a change that should address this inconsistency.

Flags: needinfo?(aflorinescu)

Thanks James,
Retested this on WIndows 10 with 116.0a1 (2023-06-12) and the issue is fixed. Could you please mark it as such, with reference to what fixed it -(was it bug 1836507 ?). Also can you clarify if we're going to uplift it to 115b - for now, 115b report is due on 06/14 and is still affected.

Flags: needinfo?(aflorinescu) → needinfo?(jteow)

Yes, it was fixed by Bug 1836507. I'll mark it as a dupe. Thanks for verifying the fix worked.

After talking with Chris, I won't ask for an uplift because the results in Nightly haven't been validated by Data Science yet, so there's good chance the experiment won't run til a post-115 release.

Status: NEW → RESOLVED
Closed: 3 years ago
Depends on: 1836507
Duplicate of bug: 1836507
Flags: needinfo?(jteow)
Resolution: --- → DUPLICATE
You need to log in before you can comment on or make changes to this bug.