Teaching an operator to use a cap gauge takes ten minutes. Making twenty operators produce the same result on the same marginal part takes a structured programme, because the decisions that vary are the ones at the edge of the tolerance band.
Assemble a set of parts with known status: clear pass, clear fail, and several marginal parts whose status has been determined by measurement. Have each operator gauge the set and compare results. This identifies both the operators who are too lenient and the ones who are too strict, which are different problems needing different correction.
Most inconsistency comes from force. Demonstrate the difference between a gauge entering under its own weight and entering under a push, using a marginal part that responds differently to the two. Operators who have seen the difference once remember it; operators who have only read the rule do not.
Retrain whenever a gauge is replaced, a sheet is amended, or a new cap family is introduced, and re-run the reference set afterwards. A short annual refresher keeps the reference set valid; without one, drift in practice is invisible until a lot is rejected by a customer.
