What is calibration? Definition, examples and why it matters

Calibration is the comparison of a measuring instrument with a reference standard of known value and uncertainty, under specified conditions, to find out how far its readings are off. That documented relationship is then used to turn the instrument’s readings into reliable measurement results. It tells you the error; fixing the error is a separate step called adjustment.

This page is about measuring instruments: gauges, calipers, thermometers, scales, pipettes and the like. It is not about monitor color calibration, car camera (ADAS) calibration or the calibration of AI models, which use the same word for different things.

What calibration means: the official definition in two steps

The reference definition is in the International Vocabulary of Metrology (VIM), published by the BIPM and the other international metrology bodies. It describes calibration as an operation with two steps, done under specified conditions (VIM 2.39):

1. Compare. Relate the values given by measurement standards, each with its uncertainty, to the corresponding indications of the instrument, also with their uncertainties. An indication is simply the value the instrument shows (VIM 4.1).
2. Use the comparison. Use that relationship to obtain a measurement result from any future indication.

Many people stop at step 1, and the VIM itself notes that the first step alone is often taken to be calibration (VIM 2.39 Note 3). Step 2 is what makes the comparison useful: once you know how the instrument behaves, a reading becomes a result you can defend.

The relationship can take several forms: a statement, a table, a curve or function, or a simple correction to add to (or multiply with) the reading, always with its uncertainty (VIM 2.39 Note 1). For example, if a pressure standard applies 300.03 psi and your gauge reads 300.93 psi, the error is +0.90 psi, and the correction is −0.90 psi: the same size with the opposite sign.

Notice what the definition does not say. It does not say the instrument must be adjusted, and it does not say the instrument passes or fails. A calibration does not “correct errors” or “minimize uncertainty”. It quantifies the error and its uncertainty, and whether that is good enough for you is a separate question, called verification (VIM 2.44). Our guide to measurement uncertainty explains where that uncertainty comes from.

Calibration vs adjustment vs verification vs validation

These four words get mixed up on shop floors, in purchase orders and in online articles. The VIM separates them clearly:

A few points behind the table:

• Calibration comes before adjustment. You can only adjust an instrument once you know its error, and after an adjustment the instrument usually has to be calibrated again (VIM 3.11 Notes 2–3). That is why certificates can show results as found (before) and as left (after).
• “Self-calibration” is an adjustment. The VIM warns against confusing calibration with adjustment, “often mistakenly called self-calibration” (VIM 2.39 Note 2). Zeroing a micrometer or running an instrument’s internal “autocal” routine is, in VIM terms, an adjustment or a correction, not a traceable calibration of the whole instrument.
• Verification is not calibration (VIM 2.44 Note 5). It checks a result against a requirement. In legal metrology, verification of a measuring instrument also includes examination and marking or a verification certificate (VIM 2.44 Note 4), which is the familiar seal on a legal-for-trade scale.
• Validation is verification where the requirement is fitness for an intended use (VIM 2.45). The VIM’s example is a measurement procedure for nitrogen in water that is also validated for human serum.

What happens in a calibration, and what you get back

Here is one test point from a typical calibration, using the 300 psi point of the digital pressure gauge in our example calibration certificate. A real calibration repeats this at several points across the range.

What you should get back from a calibration lab:

• Results with units at each test point, and the measurement uncertainty of each one. For accredited calibrations, ISO/IEC 17025 requires the uncertainty, the conditions that affect the results and a statement of metrological traceability (clause 7.8.4.1, summarized by NIST).
• As-found and as-left results when the instrument was adjusted or repaired (clause 7.8.4.1). The as-found data tells you whether the measurements you made before the calibration were good. If it is out of tolerance, that is the start of an out-of-tolerance investigation.
• A pass/fail statement usually only if one was requested (or a regulation requires it), and then with a decision rule that says how the uncertainty was taken into account (ISO/IEC 17025 clause 7.8.6, as described in Global ACI-TECH-1-002, formerly ILAC G8). Without a statement of conformity, a calibration gives you values, not a verdict. See guard banding for how labs make that call near a limit, and the TUR calculator to check whether the uncertainty is small enough for your tolerance.
• No expiry date. A certificate reports what the instrument did on one date. Labs accredited to ISO/IEC 17025 may not recommend a calibration interval on the certificate unless you agreed to it or a regulation requires it (clause 7.8.4.3).

Our guide to reading a calibration certificate walks through every field of an annotated example.

Who calibrates the calibrator? Traceability up to the SI

If your gauge is checked against a reference gauge, who checks the reference gauge? Another, better standard. The VIM calls this sequence a calibration hierarchy: a chain of calibrations from a reference down to your instrument, where each result depends on the one before it (VIM 2.40). The standards along the way have names: a reference standard is the one an organization keeps for calibrating its other standards, and a working standard is used routinely to calibrate or check instruments (VIM 5.6, 5.7).

When that chain is documented and unbroken, and each link states its uncertainty, the result of your measurement has metrological traceability (VIM 2.41). Three details matter:

• Traceability belongs to a result, not to an instrument or a lab. The VIM recommends saying “metrological traceability” because “traceability” alone also means tracking parts, samples or documents (VIM 2.41 Note 8).
• Traceable does not mean accurate enough. Traceability alone does not guarantee that the uncertainty is adequate for your use or that no mistakes were made (VIM 2.41 Note 5).
• “NIST traceable” should mean “traceable to the SI through NIST”. NIST states that it does not certify the traceability of results it did not produce, and that quoting a NIST test report number does not prove traceability (NIST traceability policy and FAQ). NIST also does not accredit calibration labs, except through its NVLAP program (NIST traceability FAQ; NIST HB 150).

At the top of the chain there is no master object anymore. Since May 20, 2019, all SI units have been defined by fixing the numerical values of seven defining constants, and those values have no uncertainty (SI Brochure, §2.2; Resolution 1 of the 26th CGPM). The BIPM describes this as possibly the biggest change to the SI since the system was created in 1960 (SI Brochure, §1). The kilogram, for example, is no longer defined by a physical artifact, the international prototype of the kilogram kept at the BIPM: it is defined by fixing the value of the Planck constant, and the kelvin by fixing the Boltzmann constant (SI Brochure, §2.3.1). National metrology institutes such as NIST realize those definitions in their labs, and every calibration below them inherits that reference.

Why calibration matters: three FDA cases

Quality standards and regulations require calibration wherever measurements are used to accept product:

• ISO 9001 (clause 7.1.5.2): when measurement traceability is required, equipment must be calibrated or verified, or both, at specified intervals or before use, against standards traceable to international or national measurement standards (ISO 9001 APG guidance). Details in our ISO 9001 clause 7.1.5 guide.
• US drug manufacturing (21 CFR 211.160(b)(4)): instruments, gauges and recorders must be calibrated at suitable intervals under a written program, and equipment that does not meet its specifications must not be used. Automatic and electronic equipment must also be calibrated or checked under a written program (21 CFR 211.68(a)).
• Medical devices: ISO 13485 clause 7.6 covers monitoring and measuring equipment, and FDA applies it through the Quality Management System Regulation since February 2, 2026 (21 CFR 820).

What happens when calibration is missing or ignored is easier to see in real cases:

• An instrument that was never calibrated. In 2018, FDA sent a warning letter to an over-the-counter drug manufacturer, Les Emballages Facoteck, because a densitometer used to calculate fill volumes had not been calibrated and had no calibration procedure (FDA Warning Letter 320-18-71, observation 3, citing 21 CFR 211.160(a) and (b)(4)).
• An out-of-tolerance result with too short a review. In 2013, FDA described a temperature recorder on a depyrogenation oven that read about 25 °C higher than the reference. The firm’s impact evaluation covered two years; FDA said it had to cover the whole period between the recorder’s last two calibrations (FDA Warning Letter to Allergy Laboratories, observation 1.b). Our out-of-tolerance guide shows how to set that look-back window.
• A reference standard that changed the results. In 2018, Roche recalled 1,163,952 boxes of CoaguChek XS PT test strips, used to measure INR in patients taking warfarin, because of abnormally high INR results. FDA classified it as a Class I recall, its most serious category (FDA recall Z-0360-2019). According to Roche’s notice published by FDA, the affected strips had been calibrated to the most recent international INR reference standard, and the replacement strips were calibrated to the previous one (Roche announcement, October 31, 2018).

The last case is a reminder that calibration is only as good as the reference it is tied to: by Roche’s own account, the strips were not “uncalibrated”; the reference they were calibrated against was what changed.

How often should you calibrate?

There is no universal interval, and no general standard says “once a year”. ISO 9001 asks for calibration at intervals you specify, or before use, without setting a duration (ISO 9001 APG guidance). ILAC G24 puts the choice on the user of the equipment, based on a risk assessment, and says no single method suits every instrument (ILAC G24, §1.2 and §4.6–4.7). ILAC G24 advises against fixed intervals unless a normative document, such as a standard or a specified method, sets them (§6.1), and regulations can set them too.

In practice, a calibration program also calibrates:

• after an adjustment (VIM 3.11 Note 3) or a repair, which is why ISO/IEC 17025 certificates report results from before and after either one (clause 7.8.4.1);
• more often while an instrument is new, until its drift is known (ILAC G24, §6.1);
• and adds intermediate checks between calibrations, compared with predetermined limits before the instrument goes back into use (ILAC G24, §4.9–4.10).

Our guide to calibration intervals shows how to set the first interval and adjust it from as-found history.

Who does the calibration?

You have three options, and most companies mix them:

• An accredited calibration lab. Labs are accredited (not certified) to ISO/IEC 17025 for a defined scope: the quantities, ranges and best uncertainties (CMCs) they can deliver. In the US, the accreditation bodies that are signatories of the Global ACI MRA for calibration are A2LA, ANAB, IAS, NAC, NVLAP and PJLA. Global ACI replaced ILAC and IAF on January 1, 2026, so you will still see the ILAC MRA mark on many certificates (Global ACI). Check that your calibration is inside the lab’s scope; our ISO/IEC 17025 guide explains how, and our lab directory will list accredited labs.
• The instrument manufacturer. Often convenient, but a manufacturer’s certificate is not automatically accredited. Review it like any other.
• In-house. Many organizations send their reference standards to an accredited lab and calibrate everyday gauges themselves against those standards. That works when the people are trained and competent, the reference standards are calibrated at suitable intervals, and the procedure and the uncertainty evaluation are documented (A2LA G149, written for accredited labs). Step 5 of our calibration program guide compares the options, and our calibration procedure guide shows how to write the method down.

For step-by-step methods for specific instruments, see pressure gauges, thermometers, scales, torque wrenches and pipettes.

Where to go next

• Setting up calibration at your company: how to set up a calibration program, then records and intervals. When a spreadsheet stops being enough, see how to choose calibration management software.
• Just received a certificate: how to read and review a calibration certificate.
• An instrument failed its calibration: out of tolerance: what to do.
• Want the math behind “±”: measurement uncertainty explained, the uncertainty calculator and the TUR calculator.
• Related basics: accuracy vs precision, what is metrology, gauge R&R (how much of your variation comes from the measurement system) and the glossary.