How a configuration record drifts: five patterns

What is drift?

Drift is the gap between what the record says and what is true. It grows quietly, a few records a week, until a team stops trusting the CMDB and starts keeping its own spreadsheet. The drift cost calculator puts a number on the manual work of closing that gap at your own counts and times.

Pattern 1: assets that never arrive

A new laptop is shipped straight to a remote employee. A developer starts cloud instances in an account nobody connected. A department subscribes to a SaaS app. None of them reach the CMDB, because nothing that feeds the CMDB saw them. This is a coverage problem, and the fix is more sources, not more effort from the service desk.

Pattern 2: ghosts that never leave

A server is decommissioned but its record stays active. A laptop is returned and wiped but still shows the last user. Ghost records inflate license counts, confuse incident routing and make audits slower. They persist because adding a record is part of a process and removing one usually is not.

Pattern 3: duplicates

The same device appears twice: once from discovery under its hostname, once from the MDM under its serial number. Neither record is wrong, but together they double-count and split the history. Duplicates are a matching problem. Lansweeper, for example, documents merging records on MAC address, serial number and hostname; lesson 6 covers identifiers in detail.

Pattern 4: stale fields

The asset exists and appears once, but a field is out of date: the operating system version, the location, the assigned user, whether the endpoint agent is still running. Stale fields are the hardest drift to see because the record looks complete. They are fixed by deciding which source is trusted for each field and refreshing from it, the subject of lesson 5.

Pattern 5: lost owners

The person who owned a server changed teams, and the record still names them. Or it never named anyone. A record without a current owner is a record nobody will correct, so this pattern makes all the others last longer.

How do the patterns map to fixes?

• Missing assets need broader discovery or more connected sources.
• Ghosts need a retire step that runs as reliably as the create step.
• Duplicates need matching rules on stable identifiers.
• Stale fields need a source of truth per field and regular refresh.
• Lost owners need ownership pulled from systems that already know, such as HR and identity directories.

The record card on each option's profile shows which of these jobs the vendor documents: found, matched, owned, synced and current.