Adobe Analytics

dymt · 9/21/25

Hi all,

I have some irregular data from Sep 17 (JST) and I think that is caused from the attached issue (https://status.adobe.com/).

The issue shows that "The Adobe engineering team reverted the data collection performance optimization change at 11:00 UTC on September 18, which prevented further unexpected values in Analysis Workspace reporting. " But I got some irregular data today (report suite and datafeed).

The irregular data looks like only post-operated data and non-post data is good for me (that is, evar1 is good, but post-evar1 is not good).

Does anyone have the same problem?

nitesh__anwani · 9/22/25

Hi @dymt ,

I have also noticed some level of data inconsistencies across all dimensions, including both custom props and out-of-the-box (OOTB) dimensions such as page_url.

When reviewing the raw data from the data feed, some values appear to belong to other Adobe clients. These values are shown as "Unknown" in Adobe Workspace.

Thanks,

Nitesh

dymt · 9/22/25

Hi @nitesh__anwani ,

thanks for your reply.

I found that post_evars in datafeed are copied from the errata data period (Sep 17 ~ 18), so the new values after Sep 19 are expected good for me (but I did not see the new value due to little events).

Best regards,

dymt

Jennifer_Dungan · 9/22/25

From all the investigation and conversations I've had and seen, this issue appears to be some sort of shifting of column data....

Even the so called "clean" data (where the data appears to be ours), is correlating incorrectly... values that only exist on our articles are showing on our home page URL for instance...

It appears to me that for the effected time of the issue, entire columns of data shifted up or down...

Basically something that looks like:

So if I was Company 2, my page 3 has someone else's value, and my page 4 has the value that should have been on my page 3... However, none of us can see which fields shifted, in what direction, or how far they shifted...

Because of this, I currently don't trust any of the data for the affected period, letting my organization know, that data in the time frame isn't currently trustworthy.

Luckily, I already had my Virtual Report Suites set up with what I call my "clean data" segment, and I added an exclusion for the hours in which the issue was active, so that no one is looking at the corrupted data.

The times are adjusted to my reporting suite time zone (which is in Eastern Time)

I will remove this after the data has been reprocessed (which I assume will take some time given the 21 hours impacting all clients (this is a lot of data to purge and re-process which new data is coming in).

It would make sense that it's the post data that is impacted, this is the data that is actually processed, and the most likely place for a shift to occur.

@nitesh__anwani when you checked your raw data feeds, were those from before the "unknown" fix was applied to try and hide garbage data, or did you re-pull the raw data feeds post this data fix?

Iceboats · 9/22/25

During the incident, I pulled a URL report from Data Warehouse and deduped to see that I was receiving data from 714 unique domains that were not our own

In Evar1 for example, I saw many unique identifiers, navigation hierarchies, search terms, and data that was possibly PII. (Was enlightening to see some of the awful news URLs and banal entertainment article titles)

Is it widely acknowledged that data from other companies was showing up across orgs?

nitesh__anwani · 9/22/25

Hi @Jennifer_Dungan ,

The raw data feeds I reviewed were loaded before Adobe applied the “unknown” fix. Since I have the ETL scheduled to run daily, it loaded the Adobe data into the data warehouse prior to the correction.

To reload the data for the affected dates, I am waiting for the official announcement from the Adobe team confirming that the issue has been fully resolved and the data is completely clean.

The overall volume of this unexpected data (coming from other Adobe clients) in the data feed is quite low. However, identifying a clear pattern is difficult—it’s unclear whether a single variable shifted, if only the “post” column was impacted, or if other factors were involved. In the raw data feeds I reviewed, I observed a mixed pattern:

Some OOTB dimensions, props, and eVars (without “post”) contained data from other clients.
Some post_eVars/props had values from my properties.
Other post_eVars/props again contained values from different clients.

Thanks,

Nitesh