Inconsistent Crash Rate Data for iOS - Possible Calculation Issue

[Yoavpagir]

Description

Description:
We are experiencing inconsistencies in how crash rates are calculated for our Mobile apps iOS environments.

For example, over the past 30 days, the overall crash rate across all releases is reported as 0.25%. However, when reviewing individual releases from the last two months, each version has a crash rate of less than 0.04%. Here

This discrepancy raises concerns that the crash rate calculation may be incorrect or affected by older versions that should not significantly impact our data.

Key Issue Identified:
Data Inconsistencies Between Total and Individual Crash Rates.

This could be due to an old versions that affecting crash rate calculation. For example, Version 2.87659.0 (released 1 year ago) still contributes to the crash rate. list of release contributing to the crash rate

1. This version only has ~150 sessions per day, which is ~0% of total sessions. Why is it affecting us so much?
2. If we look inside that version, we can see that there have been no crashes even though it mentions 1.4k crashes. Additionally, we checked and this version was generally stable for our users.

-The same behavior is observed in many other older versions.

We want to rely on accurate data for our current crash rate - session/crash in a specific environment we would like to ask for your assistance

Releases crash rate

[krystofwoldrich]

Hi @Yoavpagir,
thank you for the message and all the links,

We are looking into the data and will keep you updated here.

[krystofwoldrich]

For example, over the past 30 days, the overall crash rate across all releases is reported as 0.25%. However, when reviewing individual releases from the last two months, each version has a crash rate of less than 0.04%.

When reviewing the releases from the last two months, do you mean newly created releases? Because the Crash Rate(Session) = 0.23 includes all releases that sent data (sessions) to Sentry in the given time range (past 30 days).

---

This discrepancy raises concerns that the crash rate calculation may be incorrect or affected by older versions that should not significantly impact our data.

The combined Crash Rate includes the weight of individual Releases based on the number of sessions.

---

This could be due to an old versions that affecting crash rate calculation. For example, Version 2.87659.0 (released 1 year ago) still contributes to the crash rate.

1. This version only has ~150 sessions per day, which is ~0% of total sessions. Why is it affecting us so much?

2. If we look inside that version, we can see that there have been no crashes even though it mentions 1.4k crashes. Additionally, we checked and this version was generally stable for our users.

The reason for the discrepancy (between the number of crashes and no issues for the given releases) is the inbound filters set for the project. The events from the release shown in the screenshot are filtered out. But sessions from that release are received.

You can view the amount of filtered errors at https://{org-slug}.sentry.io/stats

You can review the filters at https://{org-slug}.sentry.io/settings/projects/{project-slug}/filters/data-filters/

---

Let us know if this helped you with understanding the data.