Data practices that hold up on a shop floor

One naming convention for plants, lines, assets, parts and metrics, applied everywhere, including tag names. Renaming later is far more expensive than agreeing now.

Check every incoming feed for schema changes, null spikes, out-of-range values, duplicate keys and missing periods. Reject or quarantine bad records rather than letting them into a metric.

Every dashboard and alert should show when its data last refreshed. A stale number that looks live is worse than no number, because it is acted on.

Route Flow failures, threshold breaches and integration warnings to the person who owns the outcome, with enough context to act. Teams that turn each recurring warning into an automation stop repeating the same fix.

Name the person accountable for each Flow, metric and app — not a department. Unowned assets rot quietly and are discovered during an audit.

Changes to a calculation must be traceable, with a date and a reason. When a number moves, the first question is always 'did the process change or did the formula change'.

Sit with the supervisor or planner and watch them use it for one shift. Everything wrong with a data product shows up in the first twenty minutes of real use.