Skip to content
CorrectiveActionReport.comSupplier CAR Response Desk — home

Fees, Scorecards & Recurrence

Repeat Nonconformance: Spotting the Same Failure Across Different Customers

A root cause that shows up on a second customer's request is often invisible until that customer escalates. This guide covers defining recurrence and counting it across every customer you supply.

Rovaryn Digital · · 12 min read

A second customer's request describes a familiar burr

A repeat is easy to miss when each request lives in a different customer's portal. Suppose one customer's complaint is a burr on a pierced hole in a bracket. The 8D is built in that customer's layout, containment goes out, the cause is traced to a worn punch, and the request closes.

Two months later a different customer reports a burr on a different part cut on the same press. The engineer who handled the first request is on another job. The new request sits in a different portal with a different complaint number. Nobody on the floor connects the two until the second customer writes "repeat issue" in the comments.

Each portal only holds one customer's complaints, and your own files may be scattered across email and spreadsheets. This article covers how to define a repeat nonconformance, which fields make repeats countable, how to count them by part, failure mode and cause, and what a supplier-side record should do about it.

What repeat nonconformance means in your own record

Customers use words like "recurrence", "repeat issue" and "repeat offender" in their manuals, and they do not all mean the same thing. Some count a repeat from the same part number. Others count from the same failure mode on the same product family. The right place to find a customer's definition is the current revision of that customer's supplier manual or CSRs, and your SQE can confirm how they apply it.

Your own record needs a definition too, because your customers' definitions will not line up with each other. If you want to know how to define recurrence nonconformity internally, start from the question you are trying to answer: "Did a cause we said we fixed come back?" That question has three parts you can write down.

  1. Same failure mode. The defect looks the same: a burr, a short shot, a porosity indication, a dimension out on the same feature.
  2. Same root cause. The cause you found matches the earlier one, not just the symptom.
  3. After closure. It occurs after the earlier request was closed, so the earlier corrective action had its chance to work.

Not every combination is the same event, and it helps to name the differences.

Situation Same failure mode Same root cause After closure What to call it
Burr returns on the same part after the fix was verified Yes Yes Yes Repeat nonconformance
Burr on a different part from the same press, same worn tooling cause Yes Yes Yes Repeat nonconformance across parts
Burr on a different part, cause is a different tool setup error Yes No Yes Similar failure, separate cause
Second complaint arrives before the first request is closed Yes Yes No Open duplicate; fold into one investigation
Same cause, different failure mode (a worn punch also causes a size shift) No Yes Yes Repeat cause, new symptom

The fourth row matters in practice. A second complaint that arrives while the first is still open belongs in the same investigation, but each customer still needs its own answer in its own format. The last row is the one most often missed, because a search on failure mode alone will not find it.

Write your definition in one short paragraph in your quality procedure and keep it separate from any customer's wording. When a customer's definition is stricter, you can still answer in their terms while counting in yours.

Why each customer's portal hides recurrence

Customer portals are channels the customer runs and the supplier must use. They do their job: they carry that customer's requests, deadlines and responses. They are not built to show you a pattern across customers, because from the customer's side there is only one supplier relationship to see.

That leaves the cross-customer view to you. The sources of history you probably have today are:

  • each customer's portal, with its own search and its own export options;
  • email threads, often held in one engineer's inbox;
  • Word files of past 8Ds, named differently by each author;
  • a spreadsheet log, if someone maintains one.

A spreadsheet or document does not connect history across customers unless someone builds and maintains that link by hand. That is a statement about how documents work, not about any person's diligence. Even a careful log goes stale when the engineer who kept it leaves or gets pulled onto a launch.

Recurrence is also something auditors look at. In simpleQuE's analysis of IATF nonconformance data, clause 10.2.3, problem solving, was the most frequent major nonconformance in IATF 16949 audits, at 10.06% of major nonconformances as of November 25, 2024. Clause 10.2.3 calls for a documented problem-solving process covering containment, root cause analysis and systemic corrective actions. That is context only. It tells you where audit attention has gone and nothing about your own results. Confirm what your certification body expects with your certification body.

Fields to capture so repeats can be counted

Nothing can be counted until each request carries the same few fields, whatever format the customer used. When you log a request, capture these in your own record in addition to whatever the customer's portal asks for. The guide to responding to a supplier corrective action request covers the intake step in more detail.

  1. Customer and customer's request number. Keep the customer's own identifier so you can find the original.
  2. Your internal ID. One ID format across all customers.
  3. Part number and part family. The family field is what lets you see a repeat across different part numbers.
  4. Feature or characteristic. The hole, the weld, the flange, the seal surface.
  5. Process step and equipment. The press, the cell, the mold, the machine.
  6. Failure mode. Chosen from a short, fixed pick list, not free text.
  7. Root cause category. Also from a fixed list (below).
  8. Occurrence or escape. Whether the cause is why the defect was made or why it was not caught.
  9. Received, contained and closed dates.
  10. Link to the earlier request, if you judge it a repeat.

Free text defeats counting. If one engineer writes "burr", another writes "burrs on hole" and a third writes "sharp edge", a search will miss two of the three. A pick list with a free-text notes box beside it gives you both.

A starting list of root cause categories

A short list works better than a long one. This one uses the familiar fishbone groupings as a starting point (ASQ lists the fishbone diagram among the seven basic quality tools):

  • Method (work instruction, setup, process parameters)
  • Machine (wear, maintenance, tooling)
  • Material (incoming lot, supplier-sourced variation)
  • Measurement (gauge, inspection method, sampling)
  • Manpower (training, instruction clarity)
  • Environment (temperature, contamination, handling)
  • Design or specification (drawing ambiguity, tolerance conflict)

Add or rename categories to fit your processes. A molder will want a mold-and-tool split, and an aerospace machining shop may want a separate fixturing category. What matters is that everyone uses the same list for at least a year, so the counts mean something.

Keep occurrence and escape separate in your tagging. A burr that was made by worn tooling and a burr that shipped because the inspection method could not see it are different causes with different fixes, and one request often has both. The guide to occurrence and escape root causes goes through how to tag them.

Counting repeats by part, failure mode and cause

Once the fields exist, counting is simple arithmetic. There are three views to build, each answering a different question.

By part and part family. Which parts keep coming back? This catches the classic case of the same part returned again and again by one customer.

By failure mode and process step. Which failure modes cluster on which equipment? This is how the burr on two customers' brackets traces to one press.

By root cause category. Which causes are we finding most often? This is where a weak corrective action shows up, because the same category keeps topping the list.

A worked example

The numbers below are invented to show the method. They are not a benchmark, and they say nothing about what your own counts should look like.

Suppose a stamper logs twelve closed requests from three customers over a year, each tagged with a failure mode and root cause category. A simple tally gives:

Failure mode Root cause category Customer A Customer B Customer C Total
Burr Machine (tool wear) 2 2 0 4
Dimension out Measurement (gauge) 1 0 1 2
Weld spatter Method (parameters) 0 1 1 2
Other (single events) Mixed 1 2 1 4

In this example, burr from tool wear accounts for four of the twelve requests, and two customers saw it. Look at that single row, and no one portal would have shown it. Customer A saw two burr complaints, Customer B saw two, and each customer might judge its own pair as unrelated to anything else.

A Pareto view (sorting the rows from most to fewest and looking at the top) is enough to decide where the next hour of investigation should go. You can do this in a spreadsheet. The work is in keeping the tags consistent, not in the chart.

Counting carefully

A few cautions keep the counts honest.

  • Count requests, not parts. One request covering a large lot is still one event for pattern purposes. Track lot size in a separate field if you want it.
  • Do not compare across periods without saying so. If you changed the pick list mid-year, note it.
  • Do not read a low count as proof of a fixed cause. A quiet period may only mean the customer has not inspected that lot yet.
  • Treat any single threshold with caution. This article does not give a recommended recurrence number, because no published benchmark applies to every shop and customer. Decide your own trigger for a deeper review, such as any second occurrence of the same cause category on the same equipment, and write it into your procedure. Check how each customer's manual treats repeats before you rely on your own trigger for that customer.

Why corrective actions fail to stay fixed

When a pattern shows up, the usual instinct is to blame the engineer or the shift. A calmer reading looks at the structure of the action. These are common ways a corrective action does not hold, described as general observations rather than findings about any one shop.

  • The cause stopped at the symptom. "Operator did not deburr" is a symptom. The 5 Whys chain should keep going to why the process allowed it.
  • Only one branch was fixed. The occurrence was corrected but the escape was not, or the reverse.
  • The fix lives in one customer's response. The change was made for the part in question but not applied to sister parts, the same process on other products, or other plants.
  • Nobody verified effectiveness. The action was implemented and the request closed, but no one checked the result after a stretch of production. The guide to verifying corrective action effectiveness covers how to plan that check.
  • Prevention was skipped. Discipline D7 exists to carry the lesson to similar processes, and under deadline pressure it is often the thinnest section. See the guide to D7 and preventing recurrence.
  • The history was not findable. Whoever worked the second request never saw the first.

The last point is about records, and it is the one a cross-customer view addresses directly. The rest are about method and discipline.

AIAG's Quality 2020 study found that problem solving was a concern across the industry. In that study, OEMs and suppliers both ranked problem solving among the two most critical issues affecting quality, according to AIAG's 2015 write-up. The survey is a decade old and general, so read it as background on why customers pay attention to repeat failures.

A repeat is rarely a surprise on the floor. It is usually a record problem first: the same cause was written down twice in two places that never met.

What a supplier-side record should do

If you are weighing a spreadsheet, a shared drive or a dedicated record for this work, these are the behaviors worth asking for. They apply whatever you choose.

  1. One record per request, across all customers. Same fields, same pick lists, regardless of which portal the request came from.
  2. Search by part, failure mode, cause and date. Not just by customer or request number.
  3. A way to link a new request to an earlier one. And to mark it as a repeat by your definition.
  4. Customer-specific clocks kept alongside. So the repeat analysis does not replace the deadline tracking.
  5. A fee and scorecard view that uses your inputs. Customer manuals and scorecards sometimes treat repeat issues differently from first occurrences. Check each customer's current revision; the overview of supplier chargebacks explains how fee terms vary.
  6. A view across plants if you run more than one. The multi-plant quality dashboard guide covers that case.

Supplier CAR Response Desk is designed to hold the supplier's own record across all of its customers. It is built so each request becomes a record with a customer, part, severity and received date, with editable per-customer response rules and a guided D1–D8 with 5 Whys and fishbone entry. The product is pre-launch, and what follows describes what it is built to do.

Cross-customer search and repeat-failure flags are designed for the Professional tier and above. They are intended to let a user search past requests across customers and see where a part or cause has appeared before. The fee-exposure view, also Professional and above, is built to calculate from the fee terms the user enters from their own customer's manual. The product does not know any customer's fees. A cross-plant view is planned for the Business tier and above.

The Response Desk records and calculates. It does not write the root cause, judge whether something is a repeat for you, certify anything, or submit to any customer portal. It is designed to be used alongside the portals your customers require, with a copy-friendly field view for re-keying into them. Pricing is one flat price per plant rather than per seat.

Whatever tool you use, the discipline above still applies: fixed pick lists, honest tagging and a written definition of recurrence.

Start with one year of closed requests

You do not need software to begin. Take your last twelve months of closed requests from every customer and add four columns: part family, failure mode, root cause category and occurrence or escape. Sort by cause category and look at the top three rows. Where the same cause turns up under two customers, open the earlier file and check whether the corrective action was verified.

If you would like to hear when the Response Desk opens, you can join the waitlist.

Put it to work

Compare plans on the pricing page.

Get the next guide

Practical guides for supplier quality teams, delivered straight to your inbox.