The prompt
Score 10 of 10, Excellent
Use it when Overtime hours or cost have gone up and you need to see where, when and why from your timesheet and workload data.
- {{audience}}
- {{period}}
- {{hours_data}}
- {{workload_data}}
- {{events}}
Full prompt and why it works
Prompt
Analyze the overtime data below and find what is driving it, for {{audience}}, who need to decide whether to change schedules, staffing or workload. The goal is to separate overtime caused by predictable workload from overtime caused by absence, scheduling or one-off events. Hours data for {{period}} (person or role, team, date or week, regular hours, overtime hours, reason code if recorded): """ {{hours_data}} """ Workload in the same period (orders, jobs, calls or units by week or day): """ {{workload_data}} """ Absences, vacancies and events (sickness, leave, open roles, breakdowns, peaks): {{events}} Produce: 1. Summary: total overtime hours, overtime as a share of total regular hours, and the change from the first half of the period to the second, with calculations. 2. Where: overtime by team or role and by week, and by day of week where the data gives dates, top 5 in each, each as a small table. Name the teams whose overtime changed most between the two halves. 3. Concentration: whether a few people carry most of it, shown as a share (for example "4 of 30 people worked 55% of all overtime hours"), without naming individuals or describing a group so small (under 5 people) that they could be identified. 4. Why: for each of the largest blocks, the explanation the data supports, such as workload above normal, cover for absence or a vacancy, or overtime on the same weekday every week that points to a scheduling pattern. Quote the figures. 5. Options to discuss: up to 4, each tied to a driver, with the hours it could affect. Rules: - Use only the data given. Do not add pay rates or costs, and do not suggest cutting anyone's hours or pay. - Treat a pattern as a lead, not proof: write "overtime rose in the same weeks as" rather than "was caused by". - Group reason codes that clearly mean the same thing, such as "cover" and "sick cover", and say which you grouped. - Keep the text outside the tables under 300 words. If reason codes are missing or unclear, say how many hours have no usable reason and analyze the rest, rather than guessing reasons.
Variables
- {{audience}}
- Who will act on the findings, for example "the operations manager and HR partner".
- {{period}}
- The weeks or months the data covers.
- {{hours_data}}
- Timesheet or payroll hours by person or role: regular hours, overtime hours, team, date, and reason if recorded.
- {{workload_data}}
- Volume over the same period, by week or day.
- {{events}}
- Absences, open roles, breakdowns or peaks during the period, or "none known".
Why it works
- What to make. It separates overtime from workload, absence, scheduling and one-off events, because each needs a different fix.
- What to avoid. No names, no groups small enough to identify people, no pay cuts, and patterns described as leads rather than causes, which keeps the analysis fair to staff.
- If unsure. Hours without a usable reason code are counted and set aside instead of being assigned a guessed reason, while obvious synonyms are grouped openly.