Dataverse long-term retention is a one-way move: choose what goes in

Power Platform apps collect business records quietly. Cases, requests, approvals, inspection results: each one is a row in Dataverse, often with notes and attachments hanging off it. After a few years the active database is full of rows nobody edits but somebody may need to find.

Dataverse has a native answer. Long-term retention moves inactive rows into a read-only retention store that stays inside Dataverse, under Dataverse security backed by Microsoft Entra ID. Learn is just as clear about the catch: once data is retained, it can't be moved back to the live application state.

That's what makes this a records decision rather than a storage tweak.

How it works, in Learn's terms

  • A maker enables long-term retention on a parent (root) table. Its child tables, attachments and images come with it.
  • A Power Platform administrator creates a retention policy on that table. The criteria is a Dataverse view. The schedule is once, daily, weekly, monthly or yearly.
  • A run takes 72 to 96 hours, and capacity reports need another 24 hours to show it. Runs consume API requests from your plan's allocation.
  • The environment must be a Managed Environment. Otherwise you can still create policies, but they're disabled.
  • Audit tables and elastic tables aren't supported.

The view is the policy

Whatever the view returns leaves the live app, together with every child row. Learn recommends a view built only for retention, tested first with a TOP N query on a small sample. We'd add a business sign-off: the app owner confirms in writing that the view selects what the policy claims, for example closed cases older than a set number of years, including their notes and tasks.

Existing logic runs too. Learn notes that delete-action cascade relationships and plugins on the table execute when a retention policy runs, and makers can write plugins for the retain action. Ask the app's maker what fires before the first production run.

Reading it back is deliberately limited

Retained data is read-only through Advanced Find, a Power Automate cloud flow, or the Dataverse OData API. Per environment, up to five users can query retained data at once, up to 100 queries a day, one table at a time, with no joins or aggregates. Results found through edit filters can't be saved or exported for sharing. For heavier analysis, Learn points to linking the environment to Microsoft Fabric, where those limits don't apply.

That suits an auditor's occasional question. It doesn't suit a team that still runs a weekly report on those rows. Find that team before the run.

Retention here isn't a Purview schedule

Dataverse long-term retention is set up in Power Apps. It isn't a Microsoft Purview retention policy, and the Learn pages describe moving rows into retention and reading them back, not a disposal schedule. Learn frames the lifecycle as active, then inactive, then deleted.

Put a named owner and a period on that last step.

Who signs

  • App owner: which rows count as inactive, and the view that selects them.
  • Records owner: whether those rows are records, how long they're kept, and who approves deletion.
  • Power Platform administrator: Managed Environment status, the policy, the schedule and the API budget.
  • Security: who gets the organization-scope read privilege needed to view retained data, and when it's taken away. Learn's own example adds an auditor to a role for the job and removes them afterwards.

Our environments and data policies post asked who may build a flow that touches a business record. This is the other end of that record's life, and it only runs one direction.

McCloy Data

Thought leadership on OpenText, SAP, and practical content and finance automation - from the McCloy Data team.

https://www.mccloydata.com
Next
Next

Core Capture for SAP historic training: choose the posted invoices it learns from