Choose a bounded reporting question
Start with a two-region table so learners can verify the result without another AI tool. State that both rows use the same unit and reporting period. The question is whether actual output met the target, not why performance changed. A small exercise makes incorrect arithmetic, reversed signs and unsupported explanations easy to see before moving to a larger workbook.
Prepare the table and the account
Use clear column headings such as Region, Target and Actual, and keep the unit and period visible. Avoid merged headings or mixing totals into the source rows. Confirm the specific Copilot in Excel experience available to the cohort. Microsoft documents assistance with analysis and formulas, but the planned exercise must be tested using the participant’s approved account and workbook setup before delivery.
Ask for a calculation and an explanation
Request absolute variance as Actual minus Target and percentage variance as that difference divided by Target. Specify how a zero target should be handled instead of accepting an unexplained error or invented percentage. Ask for the formula logic in plain language, then compare it with what the workbook actually calculates. Keep the narrative to the supplied facts; the table alone cannot establish a cause.
Verify the total rather than averaging the story
In the example, East is 12 units below target and West is nine above. The combined actual is 207 against a target of 210, a shortfall of three units or about 1.43%. Averaging the two regional percentages would produce zero and misrepresent the combined result. Ask learners to explain why the total needs its own denominator rather than an unweighted average.
Transfer the method with a reviewer
Repeat the exercise with a missing value, a zero target and a different reporting period. Learners should identify which calculations remain meaningful and which require clarification. Save the question, formulas, source table and correction notes. A finance or operations owner should approve the reporting method before it is used in a real management pack; a successful classroom example does not validate every later workbook.
