Skip to content

Platform Defaults Best Practices

Patterns and recommendations for rolling out, maintaining, and evolving Platform Defaults across your Qualytics deployment.

Set Defaults Before Onboarding New Datastores

Platform Defaults take effect at asset-creation time. Every new source datastore, container, or operation inherits the current defaults; changes made afterward do not retroactively apply unless you turn on Apply to existing datastores at save time. When you know a wave of new datastores is coming (a new team joining, a new environment, a bulk discovery from a data catalog), configure the defaults first so the assets land with the right shape from the start.

Treat Apply-to-existing as a Governance Decision

The Apply to existing datastores toggle overwrites the corresponding values on every existing datastore (and, for Observability, every non-deleted container). Per-datastore customizations that individual authors set are wiped. Because the change is a hard overwrite, treat enabling this toggle as a workspace-wide governance action:

  • Announce the change to affected teams before saving.
  • Consider a staging approach: change the global default first, wait one or two operation cycles, then re-open the tab and turn on Apply-to-existing.
  • Or run a pilot: manually update a few datastores to the new value, validate the outcome, then push the global default with propagation.

Roll Out Observability Tracking Before Automation

The cascade rule couples Tracking with AI check generation: turning off Tracking silently disables the AI check control, and enabling the AI control forces Tracking on. The safest rollout is sequential:

  1. Enable Volume Tracking or Freshness Tracking and let a scan cycle collect measurements.
  2. Review the observability charts to confirm the metrics look reasonable for your data.
  3. Return to Platform Defaults and enable Create AI volumetric checks or Create AI freshness checks.

This avoids surprise AI checks appearing on containers where the tracking measurements have not yet stabilized.

Start Profile Effort at the Default and Raise on Evidence

Qualytics ships Medium as the default AI Effort. Higher levels (High, Extra High, Max) inspect broader patterns, apply deeper analysis, and take longer per profile. Before raising the default:

  • Run one or two profiles on representative datastores at the current level.
  • Review the AI-generated checks: are they catching the right issues without false positives?
  • If your team wants finer detection, bump the default up one level at a time rather than jumping straight to Max.

Similarly, lower the Effort default only when profile time is a bottleneck and you have a plan for catching the checks that Low or Off would skip.

Match the Activate-vs-Draft Policy to Your Review Workflow

The Activate AI managed checks control decides whether AI-generated checks activate as soon as the profile completes or start as drafts that require review.

  • Activate automatically works well when your team trusts AI generation for the datastores in scope, or when the deployment is early-stage and building baseline coverage is the priority.
  • Keep as drafts works well when your team runs a review process before checks go live, when the datastores are production-critical, or when compliance requires human sign-off on new checks.

You can override the default per datastore for exceptions to your workspace-wide policy.

Choose the Right Level to Configure

Platform Defaults resolve through a precedence chain (Container or Operation > Source Datastore > Global). Pick the level that matches the scope of the decision:

  • Global (Platform Default): workspace-wide policies that apply everywhere by default. Example: your default AI Effort, your default rollup threshold.
  • Per Source Datastore: exceptions for specific datastores. Example: a legacy datastore that should not generate AI checks yet, or a high-value datastore that warrants Max Effort.
  • Per Operation: one-off decisions for a single profile or scan run. Example: a deep profile on a new dataset before you commit to a datastore-level Effort change.

If you find yourself overriding the same value on every datastore, that is a signal to change the global default instead.