Measure lifecycle, auto-cleanup and the 1,000 object ceiling

What it does

Measures are managed objects with a state, a tag set and a schedule, and the platform actively deactivates ones nobody is using. On a long-running environment the management side matters more than the builder.

Key facts

Actions on the measure list: Edit, Refresh, Rename, Activate / Deactivate, Tag, Delete, Schedule, Columns, Filter (on tags), search by name, and Disable auto cleanup.
Automated cleanup runs daily and deactivates measures that are not used in exports, in other measures, or in segments, and that were created more than 45 days ago.
Measures deactivated by cleanup are tagged SystemDeactivated, so you can find them with the tag filter.
Only a user with the administrator role can mark a measure to Disable auto cleanup.
An inactive measure reports status Skipped during a scheduled refresh rather than failing.
Active segments and active measures share a combined ceiling of 1,000 per environment; beyond that you should expect performance to degrade.
The product warns you as you approach, reach and exceed the active measure limit.
A calculated measure stored in Dataverse holding more than 10 million rows should be broken into parts.
Customer attribute measures surface in the Insights section of the individual customer profile page as per-customer KPIs; they only appear if such measures exist.
Business measures surface as KPI tiles in the Your Customer Insights section of the Home page, alongside segments and enrichments.

When to use / skip

This is not an optional feature — it is the housekeeping every environment needs after the first six months. The point at which it becomes urgent is when a client has more than one analyst building measures, because measure sprawl is the normal outcome and nobody deletes anything.

Treat Disable auto cleanup as a scalpel, not a default. It exists for measures with real business value that happen not to be wired into a segment or export yet — a board-level KPI tile, for example. Applying it broadly defeats the purpose and lands you at the 1,000 ceiling with a refresh window nobody can shorten. If a measure is genuinely important, the better answer is usually to reference it somewhere real.

Configuration decisions

Tagging taxonomy, agreed before build. Tags are the only practical way to filter and bulk-schedule measures later, and retrofitting them across a few hundred objects is grim.
Which measures get Disable auto cleanup, who approves that, and how it is reviewed.
Whether analysts get permission to create measures directly in production or work in a sandbox environment first.
The deletion policy: what happens to a measure whose owner has left, and who has authority to remove it.
Which measures are exported to Dataverse, given the 10 million row guidance and the storage that implies.
How close to the 1,000 combined ceiling the environment is allowed to run before someone is required to prune.

Gotchas

The 45-day cleanup window catches measures built during a project and only wired into activation later. The measure quietly deactivates, then reports Skipped, and the segment built on it goes stale.
A deactivated measure does not disappear from the list, so it looks fine at a glance. Check the status column and the SystemDeactivated tag.
The 1,000 limit is combined across segments and measures, so a segment-heavy environment eats the budget that measure authors assume is theirs.
Renaming a customer attribute measure changes a column name in Customer_Measure, which downstream consumers reading the Dataverse schema will notice.
Deleting a measure that is referenced by another measure or by a segment breaks the dependent object rather than being blocked cleanly, so check references first.
Only administrators can disable auto cleanup, so a delivery team without the administrator role cannot protect their own measures. Sort the role assignment out before handover.

Consultant notes

Build the measure register in week one, with a name, owner, business definition and whether it is protected from cleanup. Hand it over as a live document, not a snapshot.
Warn the client explicitly about the 45-day cleanup. It is genuinely surprising the first time it happens and reads like a platform fault.
Demo the Home page KPI tiles and the profile card Insights section to the business audience — that is where measures become visible rather than theoretical.
Push back on requests to protect every measure from cleanup. Ask instead which ones are actually used, and let the rest lapse.
Before go-live, check how many active segments and measures the environment carries against the 1,000 ceiling, and record the number in the handover so the next person has a baseline.

Worth revisiting if the cleanup window, the 1,000 combined ceiling, or the administrator requirement for disabling cleanup changes.