matomo-org / matomo

Empowering People Ethically with the leading open source alternative to Google Analytics that gives you full control over your data. Matomo lets you easily collect data from websites & apps and visualise this data and extract insights. Privacy is built-in. Liberating Web Analytics. Star us on Github? +1. And we love Pull Requests!
https://matomo.org/
GNU General Public License v3.0
19.81k stars 2.64k forks source link

Inconsistent Pageviews in Transitions Report for Same Date Range #21149

Open jmumby opened 1 year ago

jmumby commented 1 year ago

When choosing an extended date range and reviewing the transitions report, the pageview count varies when viewing between consecutive days. It's important to note that the current viewing date is not included in the selected date range.

Context

A working example would be to select a date range between 1 Jan to the 1st July. Then view the transitions report on any day after the 1st of July. The pageviews shown will be different when viewing the report on consecutive days. For example view the report, then, the following day view the same report again.

25.07.23 image

26.07.23 image

Note for shorter date ranges (less data) this doesn't seem to be an issue.

Expected Behavior

The pageviews should be the same

Current Behavior

The pageviews are different

Steps to Reproduce (for Bugs)

  1. Choose a large date range and view a transitions report for a popular page
  2. View the same report the following day (24 hours later)

Your Environment

Cloud

Stan-vw commented 1 year ago

Would be good to know if we’re showing inaccurate data here, or whether there's something else going on. @michalkleiner maybe one you could try to reproduce?

sgiehl commented 1 year ago

I actually don't understand the report. You can either select a certain period in the selector and view the transitions for this period, or you can select a certain date and vie transitions for this date. @jmumby Are you able to explain that more detailed maybe? Where are you viewing the report, and where do you select the date?

jmumby commented 1 year ago

@sgiehl I have tried to recreate the issue in demo but no luck so far. If you view this transition Then view it again tomorrow (with the same date range selected from the date picker at the top of the UI) the pageviews would be different. At the moment the results are the same and as expected in the demo. This might be because the demo has less data than the user being affected by the issue. I will try again with a wider range but it takes 24 hours to recreate the issue.

sgiehl commented 1 year ago

@jmumby If the reports for the same period in the past change the values, that can only mean that some data has been removed. As the transition reports rely on log data, it could be caused by log data being removed due to settings.

jmumby commented 1 year ago

I have tested this on the demo data, and assuming there has been no raw data deletion across this period (2022) the results for the period selected are different on different days for https://demo.matomo.cloud/index.php?module=CoreHome&action=index&idSite=1&period=day&date=yesterday#?period=year&date=2022-09-14&idSite=1&category=General_Actions&subcategory=General_Pages&popover=RowAction$3ATransitions$3Aurl$3Ahttps$3A$2F$2Fdivezone.net$2Fcayman-islands](https://demo.matomo.cloud/index.php?module=CoreHome&action=index&idSite=1&period=day&date=yesterday#?period=year&date=2022-09-14&idSite=1&category=General_Actions&subcategory=General_Pages&popover=RowAction$3ATransitions$3Aurl$3Ahttps$3A$2F$2Fdivezone.net$2Fcayman-islands

Taken September 13 2023 image

Taken September 20 2023 image

But we would need to confirm there has been no data deletion.

sgiehl commented 1 year ago

I can't see any other reason besides a possible raw data deletion or a raw data anonymization that might cause the numbers to change. If we are able to reproduce that on demo and can ensure it's not caused by data deletion we should plan in some time to investigate that in detail.