When the variance is a measurement problem

Sometimes the variance is not real. The numbers in the close do not reflect the underlying business — they reflect a measurement error. Catching these cases before they enter the variance commentary protects the operator from explaining variances that do not exist, and protects ownership from acting on data that is wrong.

Common measurement errors

How to spot measurement errors

Three signals. First: the variance is concentrated on a single date or a small number of dates rather than spread across the period. Second: the variance does not match the operational pattern (occupancy was strong but rooms revenue is below budget — the math should match). Third: the variance reverses in the subsequent period.

Any of these signals is a flag to investigate before writing the commentary. A 30-minute investigation in close week saves 4 hours of commentary writing and prevents a credibility hit when the variance reverses next month.

The reconciliation check

Build a 12-month trailing reconciliation chart for each major line item: budget, actual, variance, cumulative variance. A measurement error shows up as a sharp variance spike followed by a reverse spike — the two cancel out over 2-3 months. A real variance shows up as a sustained gap that does not reverse.

When you spot a measurement error in close, the discipline is to fix it before publishing the variance report — not to publish and correct later. Correcting later is operationally correct but politically expensive: the GM has already had the variance conversation with ownership, and "actually the number was wrong" undermines the entire reporting discipline.