Counting by weight, and when it fails

The method

Weigh a sample of known count, derive a per piece weight, and fill to a target weight with a tolerance band. For uniform stamped or wound parts this is faster and more accurate than any human count, and it is how every serious findings packer works.

Where the error comes from

Piece to piece variation, not scale resolution. A 2 mm jump ring weighs a few milligrams, so a 2 percent variation in wire diameter or plating thickness is a 2 percent count error, and the error compounds against the tolerance you promised on the label.

Which direction to err

Over. A pack that counts short generates a message, a replacement and a review; a pack with three extra jump rings generates nothing. Set the target above the stated count by enough to cover the measured variation of the specific part, not by a flat percentage.

Restate the sample when the lot changes

The per piece weight belongs to a lot, not to a part number. A new lot with a different plating run has a different per piece weight, and a checkweigher still running last month’s figure will quietly fill short all week.

Parts it does not work for

Mixed assortments, where the per piece weight is a fiction. Anything with an attached finding of different mass, such as a post earring with a pad. And parts light enough that static force is a measurable fraction of their weight, where the scale reading moves as you watch it.

What to print

Either a count with a tolerance or a weight, and be consistent about which one is the promise. Printing an exact count and shipping a checkweighed approximation is the setup for a complaint that is entirely avoidable with the word approximately.

Ground the line

A grounded fill head and a conductive chute remove most of the static that makes small plated parts both scatter and weigh inconsistently. It is the same fix for two different symptoms.

References