Establish the numbers before drafting the report.

During a fictional week, 12 requests arrive. Eight of those 12 are resolved before the cutoff and four remain open. Two older requests are also resolved. Total closures are ten, but only eight relate to requests received that week.

Write the period and each indicator’s definition. “Resolved this week” and “This week’s requests resolved” count different things. Keep definitions consistent when comparing periods.

For requests received that week, eight out of twelve are resolved: approximately 66.7%. Dividing ten by twelve would mix closures from another cohort with current arrivals. Keep numerator and denominator alongside the percentage.

Check the example by hand. In a real process, define where calculations happen and how they trace back to source records. A fluent paragraph is not verification of an operation.

Instruction: “Summarise these indicators while preserving their definitions. Separate facts and hypotheses. Attribute no change to productivity, staffing or tools without evidence. Include questions still needing answers.”

“Eight of the twelve incoming requests were closed” is supported. “AI made the team more efficient” does not follow from these data. Starting backlog is also unknown, preventing calculation of total open requests at the cutoff.

Template: period; source; indicator definitions; verified figures; description; limitations; questions; next review. If no comparable earlier period exists, state that rather than inventing a trend.

Keep a version of the data used and review numbers when filters or statuses change. The report structure can be reused, but each conclusion needs its period’s context.