Flag for review in CMDB

[Morgan Berry]

Quick question, I can see a lot of way to accomplish this, but looking to see what has worked for others.

We have come a long way in the cleanup of our CMDB, but there is always that one record, or ten, that someone sees that has data in it (so not flagged by Health checks), but the data is just incorrect. As we move toward better governance of the records, what would be a good way to accomplish the following without asking for certification/audits constantly?

1) I see a record with one or more fields with incorrect Data

2) I "flag" it for review to the Managed By Group

3) They review and make recommendations for approval/change by the Managed By or Owner.

4) I can report off of this activity.

Thoughts?

[Dr Atul G- LNG]

Hi @Morgan Berry 

Greetings!!

I am not an expert, but would like to share my thoughts:

We have come a long way in the cleanup of our CMDB, but there is always that one record, or ten, that someone sees that has data in it (so not flagged by Health checks), but the data is just incorrect. As we move toward better governance of the records, what would be a good way to accomplish the following without asking for certification/audits constantly?

1) I see a record with one or more fields with incorrect Data

2) I "flag" it for review to the Managed By Group

Atul: How you are going to flag? Like you have some style or task or mark or Excel sheet?

3) They review and make recommendations for approval/change by the Managed By or Owner.

4) I can report on this activity.

Atul: Report like list view with changes? old vs new values? For each record how many lines you will get? A tough to analyze.

Atul: Overall, if I saw, it is a more manual part, which a user / you need to do. I recommend using the Data Certificate path, and for quarterly, a round of manual checks can be done.

[Morgan Berry]

Atul, thanks, that is just it, we don't want to do certifications constantly (some people have hundreds of apps in their portfolio).  Think of this like a Knowledge feedback task, that is what we are aiming for where data can be flagged immediately and not wait for a scheduled report/certification.  Data quality relies on timeliness and even a month could have operational implications.  We could do manual, but I was looking for something that we could actually report on as part of our data quality improvement metrics (reduction in error).

We have developed a data quality model and inaccurate data, along with delays in recognizing and correcting, just hits too many boxes.  Combined with the lack of reporting just sets off my CIM/CSI alarms. 🙂

Currently, we have our CMDB open to all of IT to correct data and I can report off the number of manual changes being made.  But we want to narrow the scope of editors (as you can imagine we have had changes by one person that did not align with changes by another person) to Owned by and Managed By (editor and/or approver) and Managed By Group (Proposed change reviewers) for proper change management.

Right now what I am thinking is a button that would collect a line of text (what seems incorrect?) and create a task and assigns to the Managed By Group for review, with the actual change or approval being made by one of the other two people.  But we really try to stay OOB as much as possible, so I was looking for any other options before we put down code.

We could just add a flag and send a report daily, but that would be on the whole CI and not provide any direction to the owners.

If we had work notes or similar we could track that way, maybe.  Where someone leaves a note on a CI and the managed by group gets notified.  That would be JIT notification (and hopefully change) and we could report off the manual changes from the audit history.

Note to CMDB Team at SN... Data Quality task like a CMDB Feedback task would be a great enhancement for the CMDB workspace.  Not every piece of data is managed via Discovery. 🙂

[Ashok Sasidhara]

Based on the details which you have mentioned, it seems a data quality improvement initiative is required along with configuration management process improvements. Also data certification is needed for manual verification of key attributes. So i would recommend the following action items as the solution:

1. Configuration management process improvements: Clear roles and responsibilities should be defined about who can modify CI data. All users with ITIL and Asset roles have access to modify CMDB out of the box. It is always recommended to restrict it only to relevant users like Configuration managers, configuration analysts, CI class owners, Service owners, Application owners etc. Too many people modifying the CMDB would definitely erode the data quality.
2. Data clean up: Make use of CMDB health dashboard and CMDB & CSDM data foundation dashboards to identify the current data quality issues. Then start with initiatives to improve the data quality based on the insights from these dashboards (E.g. Remediating duplicate CIs, Analyzing the CI identification and reconciliation rules to check whether those need to be refined for any of the CI classes populated from discovery & 3rd party integrations etc.)
3. Data certification: This should be the ongoing solution for manual verification of key attributes, especially those which are populated manually (E.g. Important attributes of business services, business applications etc.)

4. Data ingestion optimization: For manual data uploads to the CMDB, make sure that the integration hub ETL is used as it follows the identification and reconciliation engine (IRE). Similarly use Service Graph Connectors for any 3rd party integrations (Again as they follow the IRE). With proper identification and reconciliation rules in place, use of these 2 methods will lead to better quality of data which is populated.

[Morgan Berry]

Ashok, yes we have all of those initiatives in place (with the edit lock-downs the reason for this question), but there is always going to be data creep from Discovery (misconfigurations) or Manually (even within a team).  By using certifications, yes, we are asking them to make reviews/changes on a schedule. 

I am looking to decrease the time that erroneous information may sit out there.  Letting it sit out there for even a month for the next certification cycle, or longer could have downstream effects to the operational teams.  Sometimes it is a typo when setting up a server, which Discovery will pick up and run with, sometimes it is the needs for a better description (in the free text fields), and sometimes it is the group/ownership information.

Those initiatives you mentioned are all great, these are all processes where you are building and mandating and on a common understanding across dozens of languages, culture and just people in general are always aligned.  They get the Technology and the Process 100%, just not the People side.  I could use queries and audits to get there, but that means know what people will do wrong in advance.  I'm good, but not that good. 🙂  And my hope is that this is rarely, if ever used, but should be an option.  Especially while companies are maturing in their CMDB/CSDM path.

So, from my perspective, let's think of the CMDB as a form of Knowledge. Would you let an erroneous (or something that is no longer correct/applicable) KB sit out there if you see the issue?  Most would contact the owner via a feedback loop.  That is what I am looking to accomplish, but in a measurable fashion as a way to see what is not getting done.  And getting this stuff out of email and into the platform.  Too many emails already.

Another example.  Let's say you have a Server record that is 100%, and you have all the Service Mapping done (does anyone really have all that?), there are still nuances to the record (or why have a Description or Comment field, right?) that require a little flexibility.  Now, your notes in the Description field have been there without issue for 3 years.  However, another BA in say Moscow needs something added, and even notices a typo in the Server name.  Now a capitalization error would probably never get picked up on a Windows system, or a heath check audit, but it is (strictly from a Data Quality perspective) inaccurate as it technically deviates from a server naming policy you have in place (even if not functionally).

But what I am getting from you and others is that there is no OOB vehicle for providing JIT feedback on CI data quality like we do for tickets, KBs, etc.  I'll give this another 24 hours and then accept your solution as accepted.  I just cannot believe a feedback loop doesn't exist somewhere for a foundational data set, where it does in almost every other part of SN.  And I just cannot believe no other DQ/CSI person has not asked in the past.  It's a head-scratcher. 🙂  

Thanks, I really do appreciate the time you took to answer.  I really do.