How can you tell whether a digital change actually worked?

Return to the original problem

Before changing a process or introducing a tool, describe what is difficult today. Is a decision delayed, an error repeatedly corrected or a customer left uncertain? Establish a reasonable baseline from real cases. It need not be a perfect dashboard to be useful.

Choose measures close to the work: how long a handover takes, how often information has to be re-entered, whether people can resolve an exception, or whether a customer receives a clear answer.

Pair numbers with observation

Usage alone can mislead. People may log in because a new process requires it while still keeping a shadow spreadsheet to get the job done. Look at representative cases, ask users what remains difficult and check whether the intended outcome is appearing.

A change can also move effort rather than remove it. If one team is faster because another now repairs more errors, the whole service has not improved.

Decide what the evidence changes

Review findings with the person who owns the outcome. Keep what helps, refine what is awkward and stop practices that add work without benefit. The purpose of measurement is to guide stewardship, not to decorate a completion report.