Test, Measurement & Calibration Services

The sector's most consequential distinction. Certification, typically to ISO 9001, says an organization operates a conforming quality management system: real, useful, and silent about whether anyone verified the measurements. Accreditation, to ISO/IEC 17025, says an independent accreditation body technically assessed the laboratory, its methods, people, uncertainty budgets, and equipment, and granted a defined scope of specific measurements at specific ranges and uncertainties. The two answer different questions: a certified lab has procedures; an accredited lab has been checked, and demanding quality systems commonly require the latter for exactly that reason. The practical test takes one minute: ask for the scope of accreditation, verify the accreditation body and number, and confirm your measurement is on the list, because accreditation certifies the list and nothing else.

As a table of what the lab may claim: parameters, ranges, and the calibration-and-measurement-capability uncertainties, the best the lab can do under ideal conditions, listed line by line; a real scope reading like mass at ten kilograms to a stated fraction of a milligram. Three habits make the reading count. Match the parameter and the range to your instrument exactly, since a lab accredited for a quantity at one range is not accredited above it, the classic audit gap. Compare the listed uncertainty to your tolerance, because a capability close to your tolerance means a thin test uncertainty ratio at your tightest points. Also check delivery (bench vs. onsite), since scopes distinguish them. The scope is public, published by the accreditation body, and reading it before the purchase order is the sector's whole buyer skill in one document.

The difference between evidence and a receipt. An accredited certificate carries the accreditation body's endorsement and the lab's verifiable number; identification of the instrument and the procedure; per-point results with as-found and as-left data; the measurement uncertainty stated point by point, not as one flattering summary; the reference standards used, with their identities and their own traceability; environmental conditions where they matter; the decision rule applied to any pass-fail statement; and the technician and date. Two traps hide in the gap: a certificate without as-found data has discarded the history you bought it to defend, and a certificate without the endorsement may be entirely honest work that your auditor cannot accept as accredited. Specify the certificate's contents in the purchase, because the paper is the product.

Traceability is a property of a measurement: an unbroken, documented chain of comparisons connecting the result back to the SI units through a national metrology institute, NIST in the United States, with stated uncertainty at every link. That is what the phrase should mean, and on an accredited certificate, it does: the standards used are listed, and their chains are auditable. What it does not mean: NIST does not approve, certify, or endorse laboratories, so NIST-traceable is not an assessment of the lab making the claim, and a calibration can be genuinely traceable from an unaccredited lab just as a sloppy one can print the phrase on anything. The buyer's translation: treat bare NIST-traceable as a claim requiring the evidence, the listed standards and their chain, and treat accreditation as the separate fact that someone independent checked the claimant.

Because calibration serves two purposes and only one looks backward. As-left data says the instrument is good for the period that begins at the sticker. As-found data, the condition on arrival before any adjustment, answers the question your quality system actually depends on: was this instrument telling the truth during the period just ended, the period whose measurements accepted product, passed inspections, and signed certificates. In-tolerance as-found quietly validates a year of work; out-of-tolerance as-found triggers the impact evaluation the OOT question below walks through. The purchasing consequences are simple: require as-found data on every certificate, require notification, not silent adjustment, when it is out, and file the data, because interval analysis, the practice of tuning recall schedules to drift history, eats as-found records and pays back in fewer calibrations of the instruments that never move.

You set the intervals: they are the instrument owner's decision, built from drift history, usage severity, environment, and the cost of a wrong measurement, with the lab's recommendation as an input. Annual-for-everything is a tradition, not a policy; interval analysis, tightening where as-found failures cluster and extending where clean history has earned it, is the grown-up version, and the metrology society's recommended practices teach it. On ratios: the test uncertainty ratio compares the tolerance being checked to the calibration's uncertainty; the trade's standing convention is 4:1, and the defense calibration standard's architecture is the clearest statement of why, capping false-accept probability at two percent with the 4:1 ratio as the fallback where that probability cannot be computed. Ask for the ratio at your tightest tolerance, and choose a stated decision rule, because if you have not chosen one, the lab's default is choosing for you.

The program's bad day, and the reason half its records exist. The lab's duty is immediate notification with the as-found data, never a quiet adjustment that destroys the evidence. Yours is the impact evaluation: what did this instrument measure during the suspect period, which of those measurements accepted product or calibrated other instruments, and does the magnitude of the error actually matter against the tolerances involved, since a small drift on a generous tolerance may close the file in an afternoon. The evaluation is only possible if reverse traceability exists: gauge IDs on inspection records, usage connected to products, which is an owner-side discipline built long before the bad day. The mature version, wired into your corrective-action system, treats each OOT as data: evidence to shorten that instrument's interval, or to retire a gauge that has failed twice running.

Same architecture, different scope grammar. In testing, the method is the product: tensile, hardness, impact, EMC emissions, vibration, thermal cycling, each per a named published method, and accreditation is held method by method, so the lab's scope reads as a standards index matched line by line against the methods your customers and regulators cite. The additions that matter: fixturing and test plans are real engineering, quoted and owned as deliverables; failure criteria are defined before the chamber closes, because a test without them is expensive noise; witnessing rights are stated where certification programs require them; and aerospace adds its own audit layer, the special-process program whose materials-testing commodity many primes flow down on top of accreditation. The routing rule is unchanged from calibration: the scope, read first, decides who can quote at all.

Standard bench calibration runs days to a couple of weeks, lab-dependent, with rush services real and priced; complex instruments, references, and method-heavy testing run longer and are quoted as projects. The logistics deserve as much planning as the dates: shipping, custody, and packaging for instruments whose whole value is their adjustment state; loaner and rental coverage for the gap, since a critical gauge in a box is a line down; and onsite service for the large, the installed, and the uptime-critical, scheduled as the plant-wide events gauge programs are built around. The program model smooths all of it, recalls scheduling spread across the year instead of a December pile-up, and one platform habit transfers whole: ask which items on your fleet list are same-week, which are send-out specialties, and let the answers, not hope, set the recall calendar.