Reporting you can check
Which revenue number answers your question?
Before rebuilding the dashboard, check what the number counts, where the clock starts and what gets left out. Here are fictional examples you can use to check your own reports.
What will someone do differently because of this number?
What decision should this report help someone make? Start there. Then check which records it counts, where the clock starts and how it calculates the result.
01 / Follow one record
The same request can produce three different answers.
In this fictional example, a request arrives at 9:00, someone creates its CRM record at 9:40 and a useful reply goes out at 10:00.
| Measure | Calculation | Result |
|---|---|---|
| How long the customer waited | 10:00 − 9:00 | 60 minutes |
| Time since the CRM record was created | 10:00 − 9:40 | 20 minutes |
| Time a person spent working on it | Observe the active work | Measure separately |
Starting the clock at CRM creation leaves out 40 minutes of the customer’s wait. Correcting that report does not save 40 minutes of staff time or make the reply faster.
Check whether the end event means an automated acknowledgment, a first human reply or an answer that actually moves the request forward.
02 / Check who is counted
Who you leave out can change the answer.
In another fictional example, 100 inquiries were due a reply. Eighty received one, 64 of them on time. Twenty are still unanswered.
64 ÷ 80 = 80% of answered inquiries met the deadline.
64 ÷ 100 = 64% of all eligible inquiries met the deadline.
Both calculations are arithmetically correct. They answer different questions. If you want to understand service across all eligible inquiries, removing unanswered work hides part of the problem.
Keep unanswered inquiries visible with their age and deadline. Agree which requests count, how to handle duplicates and business hours, and when the reporting period ends.
03 / Separate waiting from effort
Calculate staff time separately.
In this planning example, 600 requests a month each take six minutes of active work: 600 × 6 ÷ 60 = 60 hours.
You estimate that half can use an improved process and 80% of those actually will: 600 × 50% × 80% = 240 requests. Each of those takes four minutes, including review and correction. The remaining 360 still take six minutes. Allow three hours for monthly upkeep.
After: (240 × 4 + 360 × 6) ÷ 60 + 3 = 55 hours.
Potential capacity released: 60 − 55 = 5 hours a month.
That is an estimate. It depends on the volume, which requests can change, whether people use the new process and how much work remains. Five hours becomes a cash saving only if an expense falls. Include build and software costs too.
04 / Write down the definition
Write down how the report works and who checks it.
- Decision: what action should the result inform?
- Records: which ones count, including open work and difficult cases?
- Timing: what starts and stops the clock, in which timezone and business hours?
- Calculation: what is divided by what, and how do missing details affect the answer?
- Quality: what still has to be right if the work gets faster?
- Owner: who checks the definition when the process changes?
Trace ordinary, slow and difficult records back to their source. A few examples can reveal a definition problem. They do not prove how the whole business is performing.
Fix the problem the checks reveal.
You may need to correct the report’s definition, clean up records or fix how work passes between teams. Use the same checks before and after the change. A better-looking number is not enough to show better work.
A question to ask before using AI
For one deal, compare what was booked, what was billed, and what was collected. Give AI the agreed definitions and ask it to point to the records behind each answer.
What would you like to work through?
Bring a report leadership questions and the decision it should help them make. We’ll discuss what to check first.
Free, 30 minutes. Bring one example. No technical brief needed.