We have defined our research questions, measurements, experiment plan, and leading indicators in the "Add a Link - Measurement Specification". This task contains the full set of instrumentation needs, which are also copied into "instrumentation" sections on the relevant frontend tasks.
This instrumentation will exist both on the desktop and mobile versions of the feature. For users who have the Homepage enabled, we want to capture their user ID along with the information below whenever they interact with an “add a link” task.
For all measurements, we will want to:
- Distinguish by wiki.
- Analyze how much time elapses between each action.
- Know whether each action was taken from desktop or mobile.
- Know which article the user is on (through page title and page id).
- Include timestamps with millisecond precision.
The instrumentation that already tracks the user’s behavior on the newcomer homepage and in the suggested edits module should continue in the same way. The only difference should be that “add a link” should be a new task type that can be distinguished from the legacy link task.
Likewise for the help panel, which will contain guidance specific to “add a link”, the instrumentation should work the same way as for legacy suggested edits, except it will include an indicator for this new task type.
Likewise for the post-edit dialog, which should use the same instrumentation as it already does, except including an indicator for this new task type.
After a user selects a task, they are presented with three sequential onboarding screens. For each screen, we need to record:
- When the user advances back or forward through the screens.
- When the user closes the onboarding with the “Skip all” button. This event should also store whether the “Don’t show again” checkbox was selected.
- When the user closes the onboarding with the “Get started” button. This event should also store whether the “Don’t show again” checkbox was selected.
- When the user clicks the “Learn more” link.
AI suggestions mode
When the user has completed or closed onboarding, or when they have chosen the task after previously saying they don’t want to see onboarding again, they will be in AI suggestions mode ready to do the task. We need to record:
- An event for when the users arrives in the state of: they are in AI suggestions mode with the link inspector, without onboarding in front of them, either because they have completed onboarding or because it is no longer shown for them.
- It should include how many suggestions are available to them in the article at that time.
Edit mode toggle
When the user has begun the task, they are able to toggle the mode of the editor from AI suggestions to Visual Editor and Source Editor. When the user attempts to switch, they will see the modal asking if they are sure they want to switch. We need to record:
- When they open the dropdown to switch.
- The editor they select to switch to.
- When they receive the modal.
- Their button selection in the modal.
After switching, users are able to switch back to AI suggestions, though this resets them to the beginning of the suggestions for that article. We need to record an event for when they switch back.
As the user proceeds through the task, they will use the “link inspector” dialog. We need to record events for these actions:
- An impression for when the link inspector focuses on a link.
- Selecting the left arrow to go back.
- Selecting the right arrow to skip forward.
- Selecting “yes”.
- Selecting “no”.
- Opening up the blue link to view the full article of the link target.
- Reopening the rejection dialog via the ellipsis link.
- Mobile only: tapping the “?” to open the help panel.
For each of those events, we want to include:
- Which suggestion in the series they are on (e.g. “2/7”)
- The anchor text for the link suggestion.
- The target article for the link suggestion.
- The probability score for the link suggestion.
If the user changes their response on a given link, that should be recorded as an additional event, but we do not need to indicate that it was a change.
We also need to record an event for when the user selects a link suggestion in the body of the article by tapping or clicking on it.
After a user selects “no” on a suggestion, the rejection dialog opens. We need to record:
- An impression for the dialog opening.
- An event for when the dialog closes, along with the selection the user made in the list of options. We do not need separate events as the user taps around on different options -- only their final selection.
We need to record when the user attempts to advance to the edit summary screen, which can be done either through selecting the explicit “Publish/Submit” button or by advancing/skipping from the last link suggestion. When they do this, we should record which route they took.
If the user has skipped all suggestions, they’ll get the “skipped all suggestion dialog”. We should record:
- Impression for the dialog.
- An event for which button the user selects.
If the user has not skipped all suggestions, they’ll go to the edit summary screen. We need to record:
- Impression for the screen, along with counts for its contents, e.g. “7 accepted, 2 rejected, 1 skipped”
- An event when the user selects the option to “Review your changes”
- An event when the user selects the arrow to return to the article.
- An event when the user selects the Publish/Submit button.