Cycle time
The seconds a work center spends producing one piece, and the number that turns an hourly rate into a line price on the quote.
On high volume work the argument about cycle time is an argument about capital. One second of cycle at 1,200,000 pieces a year is over 330 machine hours, which means a second quoted optimistically is not a rounding error, it is a machine you may have to go out and buy. Buyers on program work know this, which is why a cycle claim that cannot survive a run at rate demonstration is a liability sitting inside the price.
The estimating failure runs in both directions. Quote the cycle from the best hour the cell ever ran and every real week loses money against the promise. Quote it from a cautious worst case and the price loses the job to whoever measured instead of guessing. The defensible cycle comes from what similar parts actually ran, averaged across real weeks with their micro-stops and tool changes included, which is a number that lives in job history rather than in a catalog.
Cycle time also decides where attention goes. On the worked spool valve in how to quote a machined part, the hone carries 5.1 seconds and $0.135 a piece, the largest inside line on the part, so a 10 percent error on the hone matters 5 times more than the same error on the wash. An estimator who knows which lines carry the seconds checks those lines hardest, and a review that treats every line equally spends its hour in the wrong places.
The quiet trap is what the cycle silently includes. Load and unload, gauging in the cycle or beside it, the chip clearing that happens every 50 pieces. Two floors can honestly measure different numbers on identical machines because they draw the boundary differently, which is why a cycle without a stated boundary starts arguments it cannot finish.