Show the Percentage of Users Affected by a Crash in a specific version on Crashlytics

[soygabimoreno]

What feature would you like to see?

I wonder if showing the percentage of users affected by a crash in a specific version is a feasible feature to implement on Crashlytics.
Sometimes is difficult to see if a crash will be an urgent issue because the adoption rate is still low.
Being able to see that would be absolutely awesome!

How would you use it?

In the day by day, to provide a better crash-free experience to our users.

Example of use case

1. A new release is uploaded.
2. The release is approved.
3. This version introduces a new crash.
4. We see a new Crashlytics issue that has triggered X crash events, affecting to Y users to the version Z of the app.

But we don't know if Y is a small or big number because we don't see if that number is big in comparison with the number of users that have adopted that version at that moment.

Currently, we have to use other tools to be able to see that proportion and that can be tough and probably can be more prone to not detecting an important crash on time.

Let's see this with numbers:
Imagine Y is 10 users.
We can think that a crash affecting to 10 users is not a big deal.
But what happens if this is only 1% of the total users in the app.
If the version that introduces the crash is on 100% of the users, that crash could be affecting 1000 users.

---

I hope the idea of the feature is clear enough.
I do think it would be a game changer and a cool companion to the Velocity alerts.

Thank you so much!

[google-oss-bot]

I couldn't figure out how to label this issue, so I've labeled it for a human to triage. Hang tight.

[argzdev]

Hey @soygabimoreno, thank you for submitting this well detailed request! This really sounds like a great idea and will help a lot of our fellow developers. I think this is something we definitely need to consider.

I’ve raised this up to our engineering team and hopefully they’ll be able to discuss this further during their team meeting sync. That being said, while I’m unable to promise any timeline for this, adding a thumbs up on this feature can help our engineers prioritize adding this to the roadmap. Thanks again!

[soygabimoreno]

Hey @soygabimoreno, thank you for submitting this well detailed request! This really sounds like a great idea and will help a lot of our fellow developers. I think this is something we definitely need to consider.

I’ve raised this up to our engineering team and hopefully they’ll be able to discuss this further during their team meeting sync. That being said, while I’m unable to promise any timeline for this, adding a thumbs up on this feature can help our engineers prioritize adding this to the roadmap. Thanks again!

Thank you very much, @argzdev 🙂

I propose an idea about how I would implement it.

I would do the same that you are doing when on hover the info (interrogation icon) to the right of Crash-free users

It shows the overall percentage filtered by the selected versions.

But...

I would add that info (interrogation icon) to each issue.

When the user on hovers that info (interrogation icon), it would show the same UI, but with the data for that specific issue.

Additionally, I would add also the percentage to be super clear.

This would be an example of getting that info from Crashlytics using BigQuery.
Take into account that the n_total_users should be fetched from Analytics.

SELECT
  issue_id,
  COUNT(DISTINCT event_id) AS `n_events`,
  COUNT(DISTINCT installation_uuid) AS `n_affected_users`,
  SAFE_MULTIPLY(SAFE_DIVIDE(COUNT(DISTINCT installation_uuid), $n_total_users), 100)  AS `rate`,
  CONCAT(blame_frame.file, ' line ', blame_frame.line) AS `crash`,
  application.display_version
FROM
  <PROJECT-NAME>
WHERE
  error_type = "FATAL" AND
  application.display_version = "$version_name"
GROUP BY
  issue_id,
  blame_frame.file,
  blame_frame.line,
  application.display_version
ORDER BY `rate` DESC
LIMIT 15

[argzdev]

Awesome stuff, @soygabimoreno! Looks like a solid plan. I'll let our engineers know about the implementation details. Thanks!