Page MenuHomePhabricator

Data request for search button
Closed, ResolvedPublic

Description

Description

In the existing search experience there are 3 ways ways to submit a search:

  1. clicking the search icon on the right
  2. clicking on a result in the dropdown menu that appears once you start typing in the search box
  3. pressing enter on your keyboard

Additionally, if you press the search icon (#1 above) without entering any text into the search box you are taken to the advanced search page (link).

In T255603 we have been discussing various options for the search button. I am wondering if we have any data on the usage of this button? I would be interested in knowing:

  • what percentage of submitted searches are submitted via the button?
  • how often is the button clicked with the search box empty?

Event Timeline

ovasileva triaged this task as Medium priority.
ovasileva raised the priority of this task from Medium to High.Aug 6 2020, 7:27 PM

Thanks for creating this task @alexhollender! We have data to answer both these questions in the SearchSatisfaction schema. I will look into this next week and let you know if I have any follow-up questions.

@alexhollender
Please see the summary of results below and feel free to reach out if you have any questions.

Data comes from SearchSatisfaction and covers the time period from January 1, 2020 through August 13, 2020. Further details available in the report.

What percentage of submitted searches are submitted via the button?
We have data available on when a user submits a search by clicking on autocomplete result in the dropdown menu or by submitting the form (which can be done either by clicking on the search icon or pressing enter). Unfortunately, current search instrumentation does not differentiate between pressing enter on your keyboard and clicking on the search icon in the widget.

A little under half (48.38%) of all search submissions using the search widget are completed by clicking on the search button (or pressing enter). The remainder of all search submissions are completed by selecting one of the autcomplete results in the drop-down menu.

submission_typenum_click_eventsprop_click_events
search_button_clicks8624120148.38%
autocomplete_clicks9201497951.62%

How often is the button clicked with the search box empty? [1]
A user clicked the search button (or pressed enter) with no input in the search box in 10.29% of all search sessions where the button was clicked and 5.54% of all search sessions.

[1] A search session does not start (and data not recorded) until a user starts to enter text into the search widget. We do not record users who click the search button with a blank text box if they have not already started a search session. After starting a session begins, if a user presses the search button again (or presses enter) with the search box empty, they are taken to the Advanced Search Page with no results shown (this is recorded in the dataset)

How often is the search box/widget used comparted to the Advanced Search Page Box?
I also took a quick look into how often users use the search box compared to the search text input provided on the Advanced Search page (link)

Only about 14% of all search submissions were completed using the text input provided on the Advanced Search Page. The rest are completed using the search box.

@MNeisler thank you so much for this report. Any ideas or options from your end on getting the breakdown between pressing enter on the keyboard and clicking the search icon?

@ovasileva @Volker_E @RHo please see the above report.

  • We now know that clicking the search button with no text in the search box happens in 5.54% of all search sessions. Hopefully the button only showing on hover (or when focusing the search field) will not negatively affect that.
  • "Unfortunately, current search instrumentation does not differentiate between pressing enter on your keyboard and clicking on the search icon in the widget." So while we know that about half the time people either press enter on their keyboards or click the search button, we don't know the breakdown within that. Perhaps we can run our own user testing and see if a similar split occurs between (1) clicking on results in the dropdown and (2) pressing enter/clicking the search icon, and then can observe the breakdown within the second category? I'm not sure if that would be sufficient.

Thanks @alexhollender for pinging on the report! I had a couple comments/questions below.

@MNeisler thank you so much for this report. Any ideas or options from your end on getting the breakdown between pressing enter on the keyboard and clicking the search icon?
@ovasileva @Volker_E @RHo please see the above report.

  • We now know that clicking the search button with no text in the search box happens in 5.54% of all search sessions. Hopefully the button only showing on hover (or when focusing the search field) will not negatively affect that.

Am I correctly interpreting @MNeisler's note [1] in T259766#6400566 that it is potentially higher than this, since sessions are not recorded if users start by clicking the search button to get to advanced search? 5-10% seems reasonably high proportion of users using the empty search as a way to 'start' their search path, so I think this helps with the decision to show the search button on hover instead of only on focus.
That said, seems the main thing would be that there is no increase in people abandoning with this change. If less people go to Advanced search but instead complete their search from the search box instead, that's a positive.

  • "Unfortunately, current search instrumentation does not differentiate between pressing enter on your keyboard and clicking on the search icon in the widget." So while we know that about half the time people either press enter on their keyboards or click the search button, we don't know the breakdown within that. Perhaps we can run our own user testing and see if a similar split occurs between (1) clicking on results in the dropdown and (2) pressing enter/clicking the search icon, and then can observe the breakdown within the second category? I'm not sure if that would be sufficient.

Agree that may not be sufficient but is still good to incorporate into testing whilst you're doing usability overall for the feature.
Is it possible to add instrumentation to figure out the breakdown?

After starting a session begins, if a user presses the search button again (or presses enter) with the search box empty, they are taken to the Advanced Search Page with no results shown (this is recorded in the dataset)

@MNeisler do we have any sense of how this happens? Is it someone typing something, deleting it, then pressing Search/enter?

Am I correctly interpreting @MNeisler's note [1] in T259766#6400566 that it is potentially higher than this, since sessions are not recorded if users start by clicking the search button to get to advanced search?

@RHo Yes that's correct. A user first needs to start typing in the search widget before we record any of their actions in our current search instrumentation so we do not account for sessions where a user starts by clicking the search button without first typing in the search widget.

@MNeisler do we have any sense of how this happens? Is it someone typing something, deleting it, then pressing Search/enter?

Yep, it looks like for search events with no text input to be recorded, a user would need to start typing something, delete it, and then press Search/enter.

Resolving this as well. Thanks @MNeisler!