1656622 - [meta] Experiment Wizard-less Password Import Autocomplete

Status: NEW

(Firefox :: Messaging System, enhancement)

Description

[Ed Lee :Mardak]

With password import autocomplete feature turned on in bug 1648182 as part of bug 1651233, we want to measure whether skipping the import wizard flow improves metrics.

This will be like bug 1618311 but with a 3rd branch:

1. no import suggestion
2. import suggestion with wizard flow
3. import suggestion without wizard flow

There are some complications that are handled by the wizard flow such as selecting browser, profile, data sources and final confirmation that won't be addressed with the skipping experiment, so we will simplify, e.g., skip wizard when there's only one profile to import. This means we'll need to have accurate telemetry of those conditions to be able to do experimental analysis.

Also based on feedback from previous experiment analysis, it might be useful to report back general sizing of how many logins to slice analysis of those with few importable passwords vs many.

Comment 1

[Matthew N. [:MattN]]

(In reply to Ed Lee :Mardak from comment #0)

3. import suggestion without wizard flow

Will it still import all data types or only logins?

If we don't offer undo then we should probably have bug 1613620 at least.

Also based on feedback from previous experiment analysis, it might be useful to report back general sizing of how many logins to slice analysis of those with few importable passwords vs many.

We have a histogram for that and if there was only one import in the session you can already assume it's what you want probably: https://mzl.la/39ODmnY

Comment 2

[Ed Lee :Mardak]

(In reply to Matthew N. [:MattN] from comment #1)

Will it still import all data types or only logins?
If we don't offer undo then we should probably have bug 1613620 at least.

jimt, any thoughts on when we should allow skipping the wizard and what it will do, e.g., all importable data from a single profile? Tim mentioned this imported data is at least additive and not destructive and into a relatively empty/new profile anyway.

We have a histogram for that and if there was only one import in the session you can already assume it's what you want probably: https://mzl.la/39ODmnY

Yeah, we used those related histograms when making observations from the previous experiment. We would need additional telemetry for experimental analysis especially for those who don't import / click the suggestion, e.g., people with few importable passwords might click the import suggestion and cancel the wizard.

Comment 3

[Jim T]

For an experiment, we don't need to support the undo / delete scenario. It may be something we revisit based on qualitative feedback.