Build a Weekly Dashboard That Does Not Lie by Accident

Your dashboard says “Leads: 40.” Does that mean forty inquiries, forty people, or forty imported rows—including duplicates? The number is only useful after the definition is clear. Start with a few metrics tied to decisions you actually make.

Choose four questions How many new inquiries arrived during the period? How many open inquiries are waiting on us now? How many next actions are overdue now? How many inquiries were closed during the period, and why? Use the inquiry tracker as the source. Notice the difference between counts across a period and a snapshot at one moment. Do not compare them as if they measure the same thing.

Write the definition beside the number Example: “New inquiries = distinct submission IDs received from Monday 00:00 to the following Monday 00:00, using the agreed business time zone; exclude test and spam records.” Define closure separately. Successful work, out-of-area requests, and duplicates should not all appear as sales. Preserve reasons using the workflow rules.

Check a small dataset by hand Create ten fictional records. Include one near the reporting boundary, one imported late, one reopened inquiry, and two requests from the same person. Count the expected results yourself, then compare the dashboard. Use Sheets filter views and Google’s pivot-table guide where appropriate. The point is to verify the calculation, not decorate it.

Keep interpretation separate “Overdue items increased from three to five” is an observation. “The team needs another employee” is a possible explanation requiring more evidence. An AI summary should receive checked numbers and instructions to distinguish observation from interpretation. Consult OpenAI’s prompting guide and apply the evidence rules from our weekly update article.

Show freshness and ownership Display the source update time and report generation time. If the source fails, show a stale-data warning instead of passing old values off as current. Give anomalies an owner through the handoff checklist. Use the debugging method to trace a wrong value to its first incorrect step, and the testing format to prevent repeat mistakes. Your dashboard is ready when you can explain every number and name the decision it supports. Use the measurement guide to check whether reviewing it improves work, rather than adding another report to maintain.