Skip to content
Ready Mark

GUIDE, read 13 Nov 2026

A readiness label goes stale the day after it is set unless something re-checks it and can take it back

The checklistAll guides

A label begins to decay the moment it is set

Three things move under a label with no one touching it: a load runs late, a test begins to fail, and a team stops using the table. Each one breaks a condition the label claimed. A label with no re-check keeps saying ready while the table underneath does not.

A schedule catches slow drift and a change trigger catches fast drift

A schedule is simple and lags by up to one interval. A change trigger reacts when a field moves but misses a table that simply goes quiet. The tools that document both are the ones to compare first. Per the vendor pages read on 13 Nov 2026:

ToolOn a scheduleOn change
OvalEdgeCertification policy yearly to hourlyRe-runs when glossary terms, description or lineage change
DataHubAt least once every dayRe-evaluated when an asset changes
OpenMetadataCertification workflows scheduled or on demandWorkflows can start on description, owner or tag changes
AtlanPlaybooks hourly, daily, weekly or monthlyNone found on 13 Nov 2026
SecodaHourly, daily, weekly or a cron expressionCan fire on schema changes
Databricks Unity CatalogOn demand or on a recurring schedule, with a dry runTags update as data and metadata change
EunoWorkflows run daily and notify by email, Slack or webhookActive tags update as the data model changes

Withdrawal is documented less often than setting

Setting a label gets a feature page. Taking it back usually does not. dbt shows a model as Degraded when a test or build fails, so the signal is withdrawn on a failed run, though dbt holds no certified status in the matrix. Databricks says its tag automations assign or remove governed tags. OvalEdge says that when a propagated Caution is removed, downstream certification reverts to the earlier status. Alation says removing a deprecation removes the flags it propagated.

The gaps are as clear. For OvalEdge, none found on 13 Nov 2026 of a step that withdraws Certified when a quality or lineage condition later fails. For Informatica, none found on 13 Nov 2026 that a certified or AI-ready label is re-evaluated on a schedule or withdrawn when a quality rule fails, though its docs site returned HTTP 403 to automated fetches. For Snowflake, none found on 13 Nov 2026 of a rule that sets or withdraws its certification tag, while the same docs say a tag removed from a source is removed from the target by propagation. They also say tags applied through CTAS, INSERT, MERGE or COPY INTO are not updated afterward.

The freshness window belongs to the use, and the check must run faster than the window

The checklist marks freshness as required for the metric agent and the write-back agent and not needed for the coding agent, so one table can be inside its window for one and outside it for another. Write the window per use in hours.

Precisely warns that a freshness check only fires if its Observer runs at least as often as the freshness rule. Apply the same test to every label: the re-check interval has to be shorter than the window. A 24 hour window checked every 24 hours can be wrong for most of a day. A practical rule is to check at no more than half the window. Other cadences matter as well. Snowflake documents data quality checks hourly by default and classification every 30 days. Monte Carlo says metadata collection is hourly, so a new tag can take time to show. Qlik says a person clicks Compute or Refresh to update dataset quality, so that signal is only as current as the last click.

A re-check checklist for every labeled asset

Use this list on a schedule and again after any incident.

  • Is the label's set date inside the review period?
  • Did the last quality run pass, and when did it run?
  • Did the last load land inside the window for this use?
  • Does the owner field still name someone who answers?
  • Is the asset still queried, and not marked deprecated?
  • Did the definition, description or lineage change since the sign-off?
  • Does the check run more often than half the window?
  • Who is told, and by what route, when any answer is no?

Write the decay terms onto the label itself

A bare "ready" flag hides all of this. A label that carries its set date, its freshness window, the date of the last check and the name of the rule or person behind it lets the agent, and the next reader, judge how old it is. Ataccama's docs show a closing pattern worth copying: freshness findings and alerts close on their own when data is written again, so the state clears without a person remembering to clear it.

Where Euno is behind

Euno does not measure data quality itself: it reads health from tools such as Monte Carlo and Elementary, no filter on its labels for agents was documented on 13 Nov 2026, and it publishes no dollar prices.

Sources

  1. OvalEdge docs: certify objects, read 13 Nov 2026
  2. DataHub docs: Metadata Tests, read 13 Nov 2026
  3. Collate docs: table certification workflow, read 13 Nov 2026
  4. Collate docs: governance workflow triggers, read 13 Nov 2026
  5. Atlan docs: playbooks, read 13 Nov 2026
  6. Secoda docs: automations, read 13 Nov 2026
  7. Databricks docs: automate tag assignment, read 13 Nov 2026
  8. Euno docs: workflows, read 13 Nov 2026
  9. dbt docs: data health signals, read 13 Nov 2026
  10. Snowflake docs: tag propagation, read 13 Nov 2026
  11. Precisely docs: freshness rules, read 13 Nov 2026
  12. Informatica docs site (HTTP 403 to automated fetch), read 13 Nov 2026

Last reviewed 13 Nov 2026.