Read the Fill Distribution Behind the Reject Count

A low reject count does not explain how much product is leaving inside each package.

On a condiment line, accepted units can still contain more material than the economic model assumes. The first task is to understand the weight measurement and the process distribution. Lowering the target comes later, if the evidence and approved requirements support it.

Reconcile what the instrument measures

Confirm whether the record is gross or net weight, how tare is established, and whether packaging variation affects the result. Check calibration, units, timestamps, and which records the export includes.

An average of accepted units excludes rejected units. It is therefore a different statistic from the average of all fills. Use the appropriate population for the question and keep both when evaluating a control change.

Reconcile instrument counts with production and reject records. Missing periods or repeated rows can create a persuasive but false picture.

Look for structure in the distribution

Separate product, line, filling head where available, batch, run stage, and approved operating settings. Choose categories that correspond to a plausible mechanism.

A high mean with a narrow spread suggests a different investigation from a broad distribution. Several narrow distributions at different means can look like one noisy process when combined. A single problematic head can disappear inside a line average.

Keep the sequence of observations. A distribution without timestamps may hide drift, changeover effects, or adjustments after a new batch arrives.

Do not assume a normal distribution or a universal number of standard deviations establishes compliance. The responsible team should determine the applicable requirements and a justified control approach.

Connect material behavior without preselecting the cause

Join batch and ingredient-lot records to the filling interval. Compare product properties measured under relevant conditions, actual settings, and weight behavior.

A supplier-lot association is a lead, not proof. Temperature, holding time, preparation, equipment condition, and measurement error may also change. A useful trial isolates the proposed cause while preserving approved operating limits.

A tighter purchased specification can be valuable if it controls a property that affects the line. It can also add cost without improving performance if the variation originates elsewhere.

Value a verified reduction

For explicitly hypothetical arithmetic, suppose a controlled, approved change reduces average net content by 2 grams across 100,000 accepted units while all acceptance requirements and output are maintained. That is 200 kilograms of reduced material use. At an assumed avoidable material cost of $3 per kilogram, the material benefit is $600 for that volume.

Those inputs are illustrative, not a benchmark. Include any change in rejects, scrap, labor, output, and other costs. Do not assume the same reduction applies to every SKU or future run.

The useful result is an explained distribution and a verified operating change. The reject count alone supplies neither.