The test was run three years ago. The engineer who ran it left the company two years ago. The customer wants to re-qualify the product for a new market that requires a different standard. Before deciding whether to run new tests or leverage the existing data, the first question is: what exactly was tested, under what conditions, with what equipment?
The test records are in a folder on a network drive called "tests 2021 final FINAL v3". Inside are a PDF of the test report (signed), a spreadsheet of raw data (format unknown, some cells with formulas that reference an external file that no longer exists), a calibration certificate for one of the sensors (the other two are not present), and a folder called "photos" containing 47 JPEG files named IMG_0001 through IMG_0047 with no captions or index.
Reconstructing what this data represents takes four days. At the end of those four days, the conclusion is that the data can probably be used, but the uncertainty about two of the test conditions is large enough that a third-party auditor would likely not accept it. New tests are run. Six weeks and €18,000 later, the re-qualification is complete.
This is not a data management problem. It is a test record problem. Here is what the records should have contained.
The minimum content of a complete environmental test record
A test record is not the same as a test report. A test report is the formal document that summarises the test results. A test record is everything that would allow someone with no prior knowledge of the test to understand exactly what was done, reproduce it if necessary, and verify that the conditions specified in the test plan were actually achieved during testing. The test record includes the test report — but it also includes the raw data, the equipment records, and the monitoring logs that the report summarises.
A complete environmental test record contains the following, in organised, retrievable form.
The test plan. Signed, dated, and at the revision that was current at the time testing began. If the test plan was revised during testing, include all revisions with the change log. This is the document against which the actual test conditions are verified. Without it, you cannot demonstrate that what was done was what was intended. The test plan requirements are covered in How to Write an Environmental Test Plan That Survives an Audit.
Sample identification and traceability. Serial numbers, batch numbers, or lot numbers for every sample tested. The hardware revision and firmware revision (if applicable) for each sample. Any pre-test conditioning applied to the samples and when. If samples were selected from a larger batch, the selection criteria. If any samples were excluded from the test or replaced during the test, the reason. Traceability from the test samples to the production population they represent — typically through the build records or quality system — is essential for the test results to be applied to the product rather than just to the specific units tested.
Equipment identification and calibration status. The chamber manufacturer, model, serial number, and asset reference for every piece of test equipment used. The calibration certificate reference, calibration date, calibration interval, and next calibration due date for every measurement instrument used — temperature sensors, humidity sensors, thermocouple reference junctions, data acquisition systems. The calibration certificates themselves, or references to where they are filed. A test record that says "calibrated equipment was used" without identifying which equipment and providing evidence of calibration status at the time of testing is not auditable.
Continuous monitoring data. The raw data from the chamber controller and any additional monitoring instruments, at a sampling interval sufficient to detect any excursions from the specified conditions. For temperature cycling: temperature at the control sensor and at any DUT thermocouples, at minimum one sample per minute. For stability studies: temperature and humidity at minimum one sample per 15 minutes. For HALT: temperature and Grms levels at the vibration table surface, at minimum one sample per 30 seconds during step stress. The data format should be human-readable without specialised software — CSV is preferable to proprietary binary formats. If the data is in a format that requires specific software to open, that software should be identified in the record and a human-readable export provided.
Deviation and excursion records. Every departure from the test plan conditions, however minor, with the date and time, the nature and magnitude of the deviation, the duration, the assessment of impact on test validity, and the action taken. The deviation handling process is covered in How to Handle Test Deviations and Chamber Excursions. If there were no deviations, the record should state this explicitly — not simply omit the section, which leaves it ambiguous whether deviations occurred or the section was forgotten.
Acceptance criteria and pass/fail determination. The acceptance criteria as defined in the test plan, the measured results for each sample at each measurement point, and the pass/fail determination for each sample against each criterion. If the acceptance criteria were met with margin, show the margin. If any criterion was not met, document the non-conformance and the disposition decision. The acceptance criteria should be in the test record even if they are also in the test plan — a reader should be able to determine whether the test was passed without having to locate and cross-reference a separate document.
Test report. Signed by the responsible engineer (or laboratory director if run by a contract laboratory), dated, and including a statement of the test conditions actually achieved (not just the specified conditions), the results, and the conclusion. If the test was run by an ISO 17025-accredited laboratory, the report should carry the laboratory's accreditation mark and scope reference.
How long to retain test records
The retention period depends on the industry and the regulatory framework. For pharmaceutical stability data: minimum 1 year after product expiry, typically longer per local regulations. For medical device qualification: the product's design history file must be retained for the lifetime of the device plus 2 years (EU MDR), or the lifetime of the device (FDA QSR 820.180). For automotive component qualification: typically 15 years after end of production (VDA 6.1, customer-specific requirements). For non-regulated electronics: there is no mandatory retention period, but the practical minimum is the product's expected service life plus 3 years — long enough to support any field failure investigation that might occur during or shortly after service life.
The retention period should be documented in the test plan and the test record. It should be enforced through the document management system, not through the memory of the engineer who ran the test.
What auditors actually check
An experienced ISO 17025 assessor or automotive customer quality engineer running a supplier audit will check the following, roughly in this order. First: can the test records be located, and are they complete? Missing documents are the most common finding. Second: are the calibration certificates current at the time of testing, and do they cover the measurement parameters and ranges used in the test? Expired calibrations and out-of-scope calibrations are the second most common finding. Third: does the raw data support the results reported in the test report? Discrepancies between raw data and summary tables are found in a significant minority of audits. Fourth: are the acceptance criteria pre-defined, or do they appear to have been written after the results were known? This is a qualitative judgment, but experienced auditors make it reliably. Fifth: is there evidence of continuous monitoring, and are any excursions documented with assessment? Gaps in monitoring data and undocumented excursions are the fourth most common finding.
The test programme management framework that connects test planning, sample selection, deviation handling, and record keeping is the subject of this cluster. The individual components — test plan, sample size, deviations, field correlation, HALT data, and records — together form a system. A programme that handles all six components correctly produces test data that survives any audit and can be used to support any decision downstream of testing.