Skip to main content

Realism and Deviations

One of DataGen’s important current design choices is the separation between:

  • enterprise richness
  • realism deviations
  • hard invariants that may not be violated

That means you can still generate a rich world without necessarily injecting the same level of flaws, omissions, and inconsistencies every time.

Deviation profiles​

The current top-level scenario control is DeviationProfile.

Available values:

  • Clean
  • Realistic
  • Aggressive

How to think about them​

Clean​

Use Clean when you want:

  • a baseline environment
  • deterministic demos with fewer distractions
  • a control world for comparison

Realistic​

Use Realistic when you want:

  • believable enterprise messiness
  • moderate drift and omissions
  • the best general-purpose default

Aggressive​

Use Aggressive when you want:

  • intentionally flawed service-management or identity views
  • harder security or discovery labs
  • more drift-heavy validation datasets

Why this matters​

Before this separation, teams often had to choose between:

  • a rich but messy environment
  • a sparse but easy-to-control one

Now the goal is to keep richness high while giving you a cleaner lever over realism intensity.

Hard invariants vs soft deviations​

Deviation profiles control the soft side of realism:

  • missing owners
  • stale CMDB views
  • conflicting policy settings
  • incomplete observed data

They do not permit hard correctness failures such as:

  • duplicate user principal names
  • structurally invalid identity records
  • impossible reference relationships

If a generated world crosses one of those hard boundaries, generation should fail instead of returning an invalid environment.

CMDB-specific override​

The CMDB profile can still carry its own override when you need the broader world at one realism level and the CMDB layer at another.