Home Power PlatformsHow to guide? Unmanaged in Dev, Managed Everywhere Else: An ALM Refresher

Unmanaged in Dev, Managed Everywhere Else: An ALM Refresher

by Mo Faheem
A courier sealing a carton box with tape before it ships

There is a moment in every Power Platform project when someone suggests making “just one quick fix” directly in production. How the team answers that sentence tells you everything about their ALM maturity. The rules of healthy solution management fit on an index card. The discipline is in never making exceptions. Here is the index card, with the reasoning attached.

Rule 1. Unmanaged in Dev, Managed Everywhere Else

Unmanaged solutions are editable working copies. Managed solutions are sealed deployments. Build in dev, export managed, import managed to test and production. A managed target environment physically resists the “quick fix in prod”, and that resistance is the feature, not the inconvenience. If production is unmanaged today, that is the first debt to pay down.

Power Apps solutions list with Unmanaged and Managed filter chips and a custom solution alongside the default solutions
The Solutions view with its Unmanaged / Managed filter chips. In dev you live on the left chip. Everywhere else, the right one. A custom solution with a real publisher sits alongside the defaults.

Rule 2. One Publisher, Real Prefixes, Boring Names

Create a proper publisher with your organization’s prefix before the first component exists, the default new_ prefix on five hundred columns is the platform equivalent of a tattoo you got on holiday: permanent, and explaining it never gets less awkward. Name solutions by scope, not by sprint: Sales Core, Service Extensions, not FixesMarchFinal2.

Rule 3. Small Solutions, Explicit Dependencies

The one-giant-solution pattern works until two teams need different release cadences, which is roughly week three. Segment by domain, keep shared components (the common tables, the base security roles) in a foundation solution that others depend on, and add components to solutions explicitly, selected components, no subcomponents you did not choose. Every accidental “add all” is a future dependency knot with your name on it.

Rule 4. Configuration Travels as Data, Settings as Environment Variables

URLs, API endpoints, feature toggles, team names, anything that differs between dev and production, belongs in environment variables, set per environment at import, never hard-coded in flows or apps. Reference data the solution needs travels in a configuration data package alongside it. A deployment that requires a human to remember to edit three flows after import is not a deployment. It is a ritual.

Rule 5. Pipelines Over Heroics

Whether you use the platform’s built-in pipelines or a full DevOps setup, the point is the same: exports, imports, and approvals happen the same way every time, logged, with a rollback story. The test is simple, could the person who deployed last month go on vacation, and could the deployment still happen this month, identically? If the answer involves a specific human and their memory, you have a hero, not a process. Heroes take holidays.

⚠️ Important: the day you find an unmanaged customization already living in production, and you will, do not build on top of it. Recreate it properly in dev, deploy managed over the top, and remove the unmanaged layer. Slow is smooth. Smooth is the only thing that survives an audit.

Closing Thoughts

ALM on the Power Platform is five boring rules applied without exception: managed targets, real publishers, small solutions, configuration as data, pipelines over heroics. None of it slows a team down after the second sprint. It is the exceptions that are expensive. Keep the index card. Decline the quick fix.

Related Articles

Leave a Comment