Diagnosing a Recurring Defect on a Packing Line
A contract packer fills 500 mL bottles of cleaning concentrate on Line 3. Customer complaints about underfilled bottles rose from 2 per month to 31 per month over eight weeks. The line runs two shifts. Quality records show fill weights averaging 498 g with a standard deviation of 4.2 g; the specification is 495-505 g. A gauge study last year found the checkweigher repeatability at 0.8 g. The night shift supervisor changed the filler nozzle two weeks before complaints began, using a spare part from a different supplier. Day-shift operators report no change in how the line behaves. The plant manager wants to recalibrate the checkweigher and issue a corrective action notice to the night shift.
1. Assess whether recalibration is the right first action. What does the data suggest about the likely cause?
Read the worked answer
Recalibration would address measurement error, but the numbers do not point there first. A checkweigher repeatability of 0.8 g is small relative to the observed 4.2 g standard deviation, and the mean of 498 g sits 3 g below the 500 g target. Measurement noise of that size cannot explain a mean shift of that magnitude or a complaint rate that rose fifteenfold.
The timing is the strongest signal. The nozzle change preceded the complaint rise by two weeks, which is consistent with a wear-in or seating period. A different supplier's nozzle may have a different orifice diameter or flow characteristic, changing the fill volume delivered per cycle.
The plant manager's proposed action also assigns blame before cause is established. Day shift reporting no change is weak evidence, because operators often normalize slow drift and because both shifts share the same filler hardware.
A better first action is to stratify complaint and fill-weight data by shift, nozzle installation date, and product batch. If the underfills cluster after the nozzle change and on both shifts, the nozzle is the leading suspect. Recalibration can proceed in parallel as a cheap check, but it should not be framed as the fix.
2. Design a short investigation that would confirm or rule out the nozzle as the cause, using only data the plant already collects.
Read the worked answer
The plant already has fill weights, complaint logs, shift records, and maintenance history. The investigation can be built from those.
First, plot individual fill weights over time with the nozzle change marked. Look for a step change or a gradual drift beginning at installation. A step suggests a dimensional difference; a drift suggests wear or seating.
Second, compare the distribution before and after, not just the mean. An increase in the lower tail matters more than a shift in the center, because underfills drive complaints. Count how many bottles fell below 495 g in each period.
Third, check whether the effect appears on both shifts. If it does, the shared hardware is implicated. If it appears only at night, the installation or setup procedure is implicated.
Fourth, compare the spare nozzle's part number and supplier documentation against the original. If orifice size differs, that alone may explain the shift.
Finally, if the data are ambiguous, swap back to a known-good nozzle for one week and compare. That is a reversible test with low cost and a clear readout.