Customer Method, Version 3 at the Bench and a Locked Spreadsheet
A small materials-testing laboratory performs tensile tests to a customer-supplied method specifying a 5 mm/min crosshead speed and a 50 mm gauge length. The method contains no precision data. Verification consisted of one operator testing one specimen at the specified speed and comparing the result with the manufacturer's typical value. When the customer revised the method last year, the laboratory's work instruction was updated to version 4, but three of the five benches still hold laminated version 3, and the electronic master on the shared drive is version 2. Results are calculated in a spreadsheet created by a former employee: one cell contains a hard-coded 0.98 correction factor with no explanatory note, and the workbook is password-protected so no one can inspect the formula. The quality manager says the spreadsheet is 'just a calculator' and needs no control.
1. Evaluate the verification of the customer-supplied method. What would a competent assessor accept as sufficient?
Read the worked answer
The verification is inadequate. For a customer-supplied method the laboratory must verify that it can properly operate the method before use - which means determining the relevant performance characteristics and confirming the method meets the customer's intended use. One operator, one specimen, one speed and a comparison against a manufacturer's typical value yields no estimate of repeatability, no reproducibility across operators or days, no bias assessment and no information about how sensitive the result is to the specified parameters.
The parameters matter. The method fixes 5 mm/min and a 50 mm gauge length; the laboratory must show it can hold them and that small departures do not swing the result. A single test at the specified speed does not answer that robustness question.
The method revision makes it worse. A revised method is a change requiring re-verification before use, and the laboratory is obliged to use the current version. Yet the verification on file, such as it is, predates the revision.
What an assessor would accept: a planned verification with defined performance criteria, several specimens across operators, days and specimen types, a bias check against a reference material or a second method, a statement of the performance achieved and its uncertainty, and records showing who approved it and on what basis. The plan can be modest for a well-established method - but it must be designed before it is executed, not reconstructed afterwards.
2. Assess the document control and data control failures. Why is the spreadsheet the more serious of the two?
Read the worked answer
Document control. Version 4 exists in a folder, the shared-drive master is version 2, and three of five benches work from laminated version 3. The laboratory cannot demonstrate which version produced any given result. Control failed at the point that matters - availability of the correct version where the work is performed - not merely in the index. The absence of a withdrawal step for superseded copies and the absence of a named owner for the work instruction are the mechanical causes.
Data control. A spreadsheet that generates reported results is part of the information system and must be validated for its intended use, protected against unauthorised or accidental change, and maintained under version control. A password-protected workbook with an undocumented 0.98 factor is the reverse of every one of those: the calculation cannot be reviewed, the factor's origin is unknown, and nobody can demonstrate that results are computed correctly. It also breaks the technical records requirement, because a calculation cannot be repeated by a reviewer.
Why the spreadsheet is worse. Document errors are visible in an audit and can be corrected by distribution. A hidden constant silently and consistently alters every reported value, in the same direction, with no trace. It is a systematic bias embedded in the system, and it defeats technical review because the reviewer sees a number, not the arithmetic. The quality manager's 'just a calculator' framing is a rationalisation: a calculator that changes reported values is precisely what requires control.
3. Set a priority order for corrective action and justify it. Then state what evidence would show the actions worked.
Read the worked answer
Priority 1: version control and spreadsheet validity. These affect results already issued, every day that passes. Actions: withdraw versions 2 and 3 from benches and the shared drive, establish a single controlled master with version, date and approver, log issue and withdrawal, and require the current version at point of use. Take the spreadsheet out of service, unlock it, document the 0.98 factor with evidence or delete it, validate the calculation against a hand calculation or reference data set, then lock it and control it as a validated tool.
Priority 2: impact assessment. Determine which work instruction version and which spreadsheet version were used for which jobs, and evaluate whether any reported result is wrong or carries unstated bias. Notify customers where results are affected.
Priority 3: method verification. Design and execute a proper verification of the current method revision with defined performance criteria. This is serious but affects confidence in results rather than silently altering them, which is why it comes third rather than first.
Root cause and effectiveness. The system cause is that document changes have no distribution, withdrawal or ownership step, and that locally developed calculation tools can enter use without review. Effectiveness evidence: an unannounced bench walk finds only current versions; a reviewer can independently recompute a reported result; the next audit samples the verification records and finds a plan, criteria and approval rather than a single data point.