Agree on the denominator
Write down which periods belong to planned production time. A lunch break with no scheduled production differs from a breakdown during a scheduled run. Apply the same rule across shifts before comparing OEE. Changes in scheduling policy can change the result even when the machine behavior is unchanged.
Reconcile counts at a physical boundary
Pick one counting boundary and verify a complete batch against the operator record. Distinguish total parts, first-pass good parts, scrap and rework. Check what happens at counter rollover, a PLC restart and product change. A counter reset should start a new interval, not create negative production.
Validate cycle time by product
Use the ideal cycle time for the product and equipment combination, with a named owner for changes. If performance exceeds 100 percent, investigate the time base and count before presenting the metric. Averaging OEE percentages from unequal shifts can mislead; retain the underlying time and count totals.
Accept a decision, not just a screen
Run one shift with a manual cross-check. Ask whether the largest reported loss matches an observed event and whether the team knows what to investigate next. Keep unclassified losses visible rather than distributing them by assumed ratios. Accept the pilot when the data supports a repeatable improvement discussion.
Put it into practice
Use the related tool to check your assumptions, then bring the results to your project discussion.