Quick tips to improve the accuracy of your metrics for deployments

One of the topics that comes up with a lot of my SCCM customers is how their management is looking at reports of things like updates deployments or software deployments and seeing low success rates.  Getting 100% success is rare in any large deployment but each company should know a reasonable number, say 90% or 95%.

For any given deployment there are many things which can and do go wrong, but there are many items which could be considered “false alarms” and really server only to bring down your metrics and distract you from focusing on the real problem clients.  I like to start by splitting all your machines into two categories: Manageable vs. non-manageable.  A non manageable machine could be corrupted, offline, or thrown away.  There are many possibilities which is its own series of discussions and leads to a discussion about the PFE client health and remediation service.

Tip 1 – Manageable machines only

There are several ways to do this:

1. For each collection add a query rule requirement so only machines having a client=1 are included
1. select SMS_R_System.Name from  SMS_R_System where SMS_R_System.Client = "1"
2. Setup your maintenance task “Clear install flag” to change this flag to no on a reasonable time frame of inactivity
2. Create a collection that is all machines where the client flag = yes
1. Limit your other software deployment collections to this collection
2. Setup your maintenance task “Clear install flag” to change this flag to no on a reasonable time frame of inactivity
3. (SCCM 2012 only) DO either of the above but use the client activity flag instead of client installed flag
1. select SMS_R_System.Name from  SMS_R_System inner join SMS_G_System_CH_ClientSummary on SMS_G_System_CH_ClientSummary.ResourceId = SMS_R_System.ResourceId where SMS_G_System_CH_ClientSummary.ClientActiveStatus = 1

The above helps you deal with machines that were manageable at one time but are no longer in a manageable state.  Of course, it is separating out the bad stuff from the good stuff in your SCCM database.  Keeping the bad stuff out of the SCCM DB in the first place is also helpful, and that means not discovering things from AD that can’t be managed.

Tip 2 – Avoid the “garbage” from AD

In SCCM 2007 or 2012

• Clean AD
• Take steps like I suggest in one of my older posts.

In SCCM 2012 only

• Filter your discovery process

Using those two tips you should be able to get most deployments hitting above 90% success.  If not then you likely have a more systemic issue like missing content or a bad command line.