A dashboard becomes useful when somebody challenges it. Why has performance improved? Which records are included? Does “completed” mean that the work was performed, inspected or invoiced? What changed in the underlying operation?
If the answer requires several people to interpret the chart differently, the reporting process needs attention before the design does.
My starting point is the management decision. A measure should help determine whether to allocate resources, investigate a problem, change a process or approve expenditure. That purpose dictates the definition and level of reliability required. A number can be accurate within one system and still be unsuitable for the decision being made.
Decide what the number represents
In a maintenance operation, requests, work orders, visits and completed tasks are different units. One request may produce several visits. A single visit may resolve several tasks. Counting them together can make activity appear to increase without demonstrating that more problems have been resolved.
The metric definition should state the unit, population, reporting period and completion rule. It should also identify the source, owner and any exclusions. These details belong in a short metric dictionary that the report preparer and operational manager both accept.
Dates deserve particular care. Work performed in one period and entered in another can distort the apparent trend. The report should explain which date governs the measure and how late entries affect previously reported results.
The UK Government Data Quality Framework distinguishes completeness, uniqueness, consistency, timeliness, validity and accuracy. Those dimensions are useful because a dataset can meet one requirement while failing another. A complete set of records can still contain duplicates or outdated information.
Examine what the chart leaves out
A performance percentage can improve because the operation improved. It can also improve because difficult work was excluded, overdue items were reclassified or the denominator changed.
For example, assume a service team closes 80 of 100 scheduled jobs. Reporting 80% completion is clear enough if the schedule and closure criteria are consistent. If 15 unfinished jobs are subsequently removed from the schedule, the same 80 completions produce a much more favourable percentage without any additional work being delivered.
That arithmetic illustrates why management needs visibility of changes to the population. Legitimate cancellations should be recorded with a reason. They should not quietly rewrite the meaning of a performance trend.
Missing information also needs its own category. An item without a target date is not necessarily on time. A blank cost is not necessarily zero. Showing these exceptions may make the dashboard less tidy, but it makes the operational position more useful.
Trace the result back to the work
Before relying on a new measure, follow selected records from the original request or transaction through the register and into the dashboard. Check whether the status, dates, quantities and classifications have survived each transfer correctly.
For significant financial totals, reconciliation should use the relevant approved records and a consistent accounting basis. Two charts agreeing with each other is weak evidence if both draw from the same incomplete extract.
The review should also test exclusions and unusual cases. A report can perform well for ordinary transactions while mishandling reopened jobs, cancelled orders, partial deliveries or corrections. These cases often matter precisely because management is interested in exceptions.
Once the measure is in use, changes to its definition should be controlled. A new completion rule may be sensible, but the historical comparison needs to explain the break. Otherwise, a change in reporting can be presented as a change in performance.
Keep comparisons commercially fair
A comparison between properties or teams should reflect the work being performed. Cost per room, hours per job or revenue per employee may be useful, but differences in service scope, occupancy, condition and work complexity can affect the interpretation.
Ratios should be shown with the underlying quantities when they inform a material decision. A small sample can produce a dramatic percentage change. A large absolute backlog can remain significant even when the completion rate looks strong.
The report should distinguish an observation from its explanation. A decline in response time is an observation. Attributing it to a staffing change requires evidence. Automated commentary, including AI-generated summaries, should be held to the same standard.
The final responsibility should be clear: who owns the source record, who checks the report and who can authorise a correction? When an error is found, correct the underlying data and identify whether previous decisions or reports were affected.
A strong dashboard does not eliminate questions. It makes better questions possible and provides a dependable route to answering them. That is the standard against which I would judge the investment in management reporting.