Knitted Bag Corrective Action Process: Containment in 24 Hours, Root Cause by Five Whys, and Closure Verified Over Three Lots

What a Corrective Action Process Is For

A corrective action exists for one reason: to make a specific defect stop happening. Not to explain it, not to apportion cost, and not to produce a document. A process that ends with a satisfactory explanation and a reworked lot has failed, because the defect will return on the next run and the buyer will pay for it twice.

The distinction that matters is between correction and corrective action. Correction is sorting the bad pieces out of this lot. Corrective action is changing something so the next lot does not contain them. Most of what passes for corrective action in bag programmes is correction, which is why the same defects appear season after season on otherwise well-run programmes.

The eight-discipline format, usually shortened to 8D, is the standard structure and it works well for a knitted bag programme because it forces the two steps most often skipped: root cause and verification. Both are uncomfortable, root cause because it often lands on a decision made months earlier, and verification because it costs time after everyone wants the file closed.

Step Zero: Containment Within 24 Hours

Containment happens before the investigation and it is not negotiable. Within 24 hours of a rejection, three things must be true: the affected stock is physically identified, it is segregated from stock that is not affected, and it cannot ship by accident. Everything else can wait a day; those three cannot.

Identification means knowing the scope, and scope is defined by the lot, not by the carton. If a gauge drift was found on one machine, the scope is every piece knitted on that machine since the last verified check, which may be several lots and may already be packed. The temptation to narrow the scope to the rejected lot is strong and it is how a defect reaches a customer twice.

Segregation is physical rather than administrative. A pallet with a label saying hold is not segregated, because labels fall off and night shifts move pallets. A cage, a marked floor area with a barrier, or a separate room is segregation, and the containment record should state which was used and who holds the key.

The Eight Steps, With an Owner and a Date

Each of the eight steps needs a named owner and a date, and every step without both of those will be the step that quietly does not happen. The table sets out the eight with the output each one produces, because an output is what makes a step verifiable rather than merely claimed.

D4 is the step that decides whether the process works, and it asks two questions rather than one: why did the defect occur, and why was it not detected before it reached the customer. The second question is the one that produces durable change, because it usually exposes a missing check rather than a machine fault.

Root Cause: Five Whys on a Knitted Line

Five Whys is a blunt instrument and it works, provided the answer to each why is a fact rather than a category. The failure mode is answering with a department name or a person, which stops the chain immediately and produces a corrective action of tell them to be more careful, which is not a corrective action at all.

1. Why are there dropped stitch runs on the outer face? Because yarn tension varied at feeders 6 and 7 during the second shift.
2. Why did tension vary? Because the tension discs were cleaned at shift change and re-set by feel rather than to the run sheet.
3. Why were they re-set by feel? Because the run sheet records tension at start-up only and not the setting after cleaning.
4. Why does the run sheet not record it? Because the cleaning procedure was added to the line without a corresponding change to the document.
5. Why was the procedure added without the document change? Because changes to cleaning routines are not routed through document control.
6. Root cause: cleaning procedures can be changed without a document control review. Corrective action: route all routine changes through document control and add a post-cleaning tension setting to the run sheet.

Notice where that chain ends. The obvious corrective action after the first why is re-tension the feeders, which fixes today and changes nothing. The chain ends five levels down at a document control gap, and fixing that prevents every future instance of the same class of problem, not just dropped stitches on this run.

The same discipline applies to the detection question. Why was it not caught? Because the in-process check samples one panel per hour. Why one per hour? Because that was the rate when the line ran one style. Why has it not changed? And so on, usually to a review trigger that was never written.

Choosing a Corrective Action That Survives

Not every action is equal, and the ranking is consistent. The strongest actions remove the possibility of the error, the next strongest make it visible immediately, and the weakest depend on someone remembering. Actions in the third group are the ones that fail within a season, usually after a staff change.

The rebound check in the fifth card is worth stating as a requirement. Tightening a yarn feed to fix an abrasion result will change loop length, which changes weight, hand and shrinkage, and a corrective action that silently moves three other properties has replaced one problem with three.

Verification: Three Lots, Not One

Verification on a single lot proves very little, because one good lot after a change is what you would expect whether the change worked or not. Three consecutive lots without the defect is the working standard, and it is the minimum that distinguishes a real effect from normal variation.

The verification must use the measurement that failed, not a general inspection pass. If the rejection was a dimensional result, verification is five measurements per colourway per lot showing the mean inside half the tolerance band. If it was an attribute count, verification is the count on all three lots with the defect class and the numbers recorded.

Where three lots would take too long, the alternative is a defined interim: 100 percent screening with the screening results reported lot by lot, ending when three consecutive screened lots show the defect below an agreed rate. That is slower and more expensive than a real fix and saying so out loud is usually what motivates finding one.

What to Record and What to Keep

The record is what makes a corrective action reusable, and it is also what a retailer asks for when a recurring problem reaches them. Nine fields cover it, and the file should be searchable by defect class rather than by date, because the question a buyer eventually asks is whether this has happened before.

Defect class and trigger

The catalogue line the defect failed against, with its version number

Scope

Lots affected, quantity contained, quantity already shipped

Containment action

What was physically segregated, where, and by whom

Problem description

Measurable: what, where, how many, since when, with photographs

Root cause statement

Survives five Whys, plus a separate statement on the detection gap

Corrective action

The change made, with the date implemented and the person responsible

Verification results

Measurements or counts from three consecutive lots

Documents revised

Tech pack, QC plan and defect catalogue, with new version numbers

Closure

Signed note, date, and any cost settlement agreed between the parties

Retention is the field buyers forget. Keeping the file for the life of the programme plus a season means that when a similar defect appears eighteen months later, the first question can be answered in minutes rather than by re-running an investigation that has already been run once.

Pareto: Which Defects Deserve a Full Process

Not every defect warrants eight disciplines. Running the full process on every finding exhausts the team and dilutes attention, and the standard filter is a Pareto: run the full process on the defects that account for the bulk of the cost, and handle the rest with a simpler correction and a watch list.

In practice that means the top two or three defect classes by frequency or by cost, which on bag programmes usually account for around 80 percent of the total. Everything below the line gets recorded, monitored and reviewed monthly, and anything that climbs the list gets promoted to a full process.

The promotion rule is what makes the filter safe. A defect that is rare this season and common next season will be caught by the monthly review rather than by the initial ranking, provided the review exists and somebody owns it. Both of those need to be written into the plan or neither happens.

Closing the Loop Into the Next Programme

The final step is the one that converts a corrective action from a cost into an asset: feed the lessons into the next programme's specification before it is written. A revised tolerance, an added in-process check, or a different yarn that came out of an 8D should appear in the next tech pack automatically rather than when someone remembers.

Frequently Asked Questions

People Also Ask