Thanks for sharing your knowledge. That makes sense because Cognos would generate the SQL based on the recent version and every version of every programming language has some changes in the way it gets compiled by the compiler and subsequently executed by the database. I have spent quite a bit of time comparing the SQLs generated for the same report in different versions and found some changes.

I think the reports should be upgraded as a best practice, to inherit the newer features and to prevent them from stop working. 

My experience has been that report upgrades are typically not needed.  But I like to get new features -- like moving from 10 to 11 and enabling interactivity -- so that my users can have more streamlined workflows.  I also converted as much as I reasonably could from CQM to DQM to prepare for a possible move to the IBM Cloud.  One thing I have found is that any of these upgrades can cause problems.  Reports must be reviewed.  In my environment, with 12,000 reports, 3000 distinct users total, 250 distinct users per day, and hundreds or thousands of report runs per day, I let users do the "testing" in the normal course of their duties.  This works for me because I find that most report problems can be corrected quickly.  (Keep in mind that this is in the context of system upgrades.  Naturally, if we're making database changes, reports get tested before inflicting them on users.)

One thing I have seen -- and I think this relates to DQM -- is that performing multiple layers of calculations in a query can get a little sketchy.  An example is if you create a data item that summarizes another data item that is calculated from a third one that is sourced from an upstream query and use it as a filter.  While this type of thing worked flawlessly for years, I have found that some are failing in recent Cognos versions.  The solution is to break the steps into separate queries (Query1 -> Query2 -> Query3) to get the aggregation correct at each stage.

Hello everybody
We use IBM Cognos v.11.1.7
We have a main goal to optimize the Enterprise report system by dropping or editing schedules with very high frequency or with very large periods of dates. Now in our system there are 3140 schedules created by business-users. The original reports are located in different folders. An usually business-users make a copy of original report and save it in their own folders, and then create a schedule for that copy
For that purpose we use MotioPI third-part application - to found the schedule, its creator and the original name of the report, from which the copy was made
But recently we faced a problem with "deserialized object" - when the xml of the object is getting invalid and third-part application cannot open it (while the native Cognos apps can open it without any problem)
here is the description of the problem: https://support.motio.com/portal/en/kb/articles/motiopi-error

The question is - what can cause this error and what can we do to prevent it?

Depends on the skills you have and what you are looking for. I see lot of job ads including ads allowing remote work. Definitely more opportunities than an year ago. Looks like Cognos is coming back up though Power BI still dominates. 

Found a good course from Texas A&M University. 6 month fully remote course, can be done in your own time and covers Python, ML and other things for becoming an AI developer. Going to start on it soon

I know this is a dead forum but posting this just in case..

I saw today's demo by Rachel Su ad wanted to ask if anyone has seen Agentic AI make metadata models and/or reports. She did not show that in the demo and I wonder if that AI can actually make reports without the need of a report developer? As everyone knows, most organizations do not have a self-service BI model and the packages they have allows only to pull the data which requires a developer. Can AI just take the data from the database and do the rest to be able to make the reports? Can AI make the metadata models from an ERP schema which is in the snowflake design?

The generative AI I have seen so far in google etc. scans the sites and interpretes the content like a human brain..

Quote from: dougp on 15 Dec 2025 10:20:30 AMThat's always a good double-check.  So if you write custom SQL for this, you get different results than what Cognos is producing?  Cognos is just generating SQL for you based on what you have created in Reporting combined with the way the model was defined.  So what's different between the SQL that Cognos is generating and the SQL that you wrote that produced the result you wanted?  Did you make a mistake in the report?  Is there an incorrect relationship in the model?  ...?

You provided a very simple query as an example but the results are different in case of complex and/or multiple queries. With inner and outer queries you can control which filter is applied where and make sure they are exclusive which is what the OP's problem seems to be. Many people create multiple filters without realizing that they are not exclusive and one filter will clash with the other causing the results to be wrong. It is true that Cognos will still re-write the SQL you write but it does not alter the logic of your SQL. I am working for a company which used Deltek's  Costpoint which is an ERP only for government contractors (project based sales only) so the data is small but there are hundred of reports which take hours and even days to run. I re-made those reports using hand written SQL and they run in seconds now! If any report is using joins from a package then never create joins in the report. It's better to write your own SQL which is a complete SQL containing all queries and use that complete SQL in one query in the report instead of creating separate small SQLs and then joining them in the report. That's what did the trick.

There is a company where the BI manager always installs the latest version of Cognos as soon as it is released but never upgrades the reports. He didn't even know that reports should be upgraded until I told him. Most of the reports are written in version 8x and 10x. While IBM says that they would run in newer versions even if not upgraded but this time I saw some reports stopped running showing syntax errors in the latest version while the same reports are running in the prior version.

Any thoughts? It maybe that the SQL generated in the latest version is slightly different than the earlier one.

How did you end up solving this?  I'm looking for a way to make Cognos available through the same authentication we use for M-365, which I understand is SAML-based.

This topic has been moved to Reporting.

https://www.cognoise.com/index.php?topic=38483.0