How to write a calibration procedure (with an example and template)

A calibration procedure is the written, step-by-step method for calibrating one type of instrument: which parameters and ranges to check, with which reference standards, under what conditions, at which test points, how to decide pass or fail, and what to record. It has to be detailed enough for a qualified technician to carry out the calibration properly and in the same way every time (VIM 2.6; NASA-HDBK-8739.19-2, §10.2.1). It is not your calibration program, which covers the whole inventory, intervals and out-of-tolerance handling, and it is not the form you fill in. This guide covers what goes in each section, how to choose test points and acceptance criteria, as-found and as-left data, environmental conditions, method verification and validation under ISO/IEC 17025, and an annotated example with a printable checklist.

Calibration procedure vs program, work instruction and record form

These four documents get mixed up, and many “calibration SOP” templates online are actually program procedures or record forms. No standard defines all four terms. These are the working definitions we use across our guides:

Regulators expect both layers. FDA’s drug CGMP rule requires a written calibration program with specific directions, schedules, limits for accuracy and precision, and remedial action (21 CFR 211.160(b)(4)). MDSAP auditors accept a general program procedure plus more specific procedures for calibrating individual equipment (MDSAP Task 14, p. 108). In ISO/IEC 17025, a “method” can be read as a measurement procedure in the VIM sense (note to clause 7.2.1.1, as quoted by PJLA). For the program layer, see our guide to setting up a calibration program.

Do you need a written calibration procedure?

In practice, yes, if you calibrate anything in-house. The requirement is worded differently in each framework. For paid standards we describe requirements in our own words, based on the free sources cited.

Our ISO 9001 clause 7.1.5 guide and ISO/IEC 17025 guide cover the rest of each standard.

What a calibration procedure should contain

No official template exists. The structure below is our synthesis of four free sources: NASA’s list of what a laboratory calibration procedure typically includes (NASA-HDBK-8739.19-2, §10.2.1), EPA’s guidance on writing technical SOPs (EPA QA/G-6, §3–4), the optional Z540-1 requirements in NVLAP’s handbook (NIST HB 150-2, Annex A.3), and the layout of NIST’s own published procedures (NISTIR 6969, SOP 4: introduction and prerequisites, methodology, calculations, measurement assurance, uncertainty, certificate). NCSL International also publishes a recommended practice on calibration procedures (RP-3), which NASA cites for guidance; it is a paid document and we have not used it here.

How much detail? EPA’s test is useful: someone with basic knowledge of the work but little experience should be able to follow it without supervision (EPA §2.1). The right level depends on how critical the measurement is, how often the procedure is used, how many people use it and how much training they get (EPA §3.0). NASA sets the same bar for in-situ procedures as for laboratory ones (§10.2.2).

Example: an annotated calibration procedure for a digital pressure gauge

The example below is an in-house procedure for digital pressure gauges of the same type as the gauge in our example calibration certificate and calibration record: 0–300 psi, 0.01 psi resolution, tolerance ±0.75 psi (±0.25 % of full scale). It is condensed onto one page; a real procedure would run to three or four. Everything in it is fictitious. Select a number to jump to its note.

Three choices make this procedure hold up. The reference standard is specified by performance, so replacing it doesn’t require a new revision. The steps force as-found data first and only allow an adjustment past a written threshold. And the target uncertainty is tied to the decision rule: if the calibration can’t reach U ≤ 0.18 psi, a simple pass/fail against ±0.75 psi is no longer valid, and the procedure says what to do instead.

How to choose test points

Test points decide what the calibration can prove. A gauge checked only at zero and full scale tells you nothing about the middle of the range, where it may be used every day. The sources agree on the principle and differ in effort:

Three more rules from the same sources:

• Calibrate it the way it is used. If the instrument works with an accessory or fixture, calibrate it in that configuration (EURAMET cg-15, §3.3.7), and respect the mounting position (DKD-R 6-1, §7).
• Calibrate at use conditions when they matter. Equipment used at process temperature may need calibration at that temperature (NASA §10.2.1).
• Cover what you actually measure. If a caliper is used mostly at 150 mm, a test point near 150 mm is worth more than one at 10 mm. A caliper that reads zero with the jaws closed has been checked at one point, not calibrated over its range (NASA §10.2.1).

Acceptance criteria: tolerance, TUR and decision rule

A procedure without acceptance criteria produces numbers, not decisions. It should state the range and the tolerance or uncertainty for each parameter calibrated (a requirement under Z540-1: NIST HB 150-2, Annex A.3.1) and the performance criteria for deciding conformance (NASA §10.2). FDA’s drug rule asks for limits for accuracy and precision in the written program (21 CFR 211.160(b)(4)).

Where the tolerance comes from. Pick one source and write it down next to the number:

• the manufacturer’s specification (the default for most instruments);
• a product standard for that instrument type, such as ASME B89.1.14 for calipers, ISO 6789 for torque tools or ISO 8655 for piston pipettes;
• a process requirement: a tighter or wider limit derived from what the instrument measures. If you set a wider limit than the manufacturer’s, restrict the instrument’s use to work where that limit is good enough, and label it.

How uncertainty is handled at the limits. If the procedure leads to a pass or fail statement, ISO/IEC 17025 expects a decision rule that takes measurement uncertainty into account, documented and, for customers, agreed (Global ACI-TECH-1-002, which replaced ILAC G8 in 2026 with the same content). The choice is between simple acceptance, guarded acceptance with a guard band, and non-binary rules. Our guard banding guide explains each one.

The TUR. Many procedures require a minimum test uncertainty ratio, often 4:1, and allow simple acceptance only above it. That is a sound way to write it, as long as you know what it is: A2LA calls a TUR of 4:1 or greater the generally accepted industry standard (A2LA G136). Use the TUR calculator to check yours.

Reference standards and equipment

Describe what the reference standard must achieve, not just which one you own: the parameter, range, and tolerance or uncertainty, optionally with representative models (NIST HB 150-2, Annex A.3.2). US Army calibration procedures show the format well: a table of minimum equipment specifications chosen to give a 4:1 ratio, with the actual accuracy stated where 4:1 can’t be reached, and permission to use alternatives that meet the minimum and are in calibration (TB 9-6625-2372-24, §4). That is Army practice, not a general requirement, but it keeps the procedure valid when equipment changes.

NIST’s mass procedures add the prerequisites that belong at the start of any procedure: valid calibration certificates with values and uncertainties for every standard, traceability to the SI, standard uncertainties small enough for the level of calibration, and measuring equipment in good operating condition (NISTIR 6969, SOP 4, §1). ISO 9001 states the traceability part for any organization where measurement traceability is required: calibration or verification against standards traceable to international or national measurement standards, and a record of the basis used where none exist (clause 7.1.5.2).

As found, adjustment and as left

Calibration and adjustment are different operations. Calibration compares the instrument’s indications with reference values; adjustment changes the instrument so it reads closer to them, and after an adjustment the instrument normally has to be calibrated again (VIM 2.39 Note 2; 3.11 Note 3). When a procedure includes adjustments, the organization should keep records of the as-found and as-left conditions, and it is good practice to keep the measured values with their uncertainty (NASA §10.2). Calibration certificates must report results before and after adjustment or repair, where available (ISO/IEC 17025 clause 7.8.4.1), and ANAB expects before-and-after data even from non-accredited suppliers in the traceability chain (ANAB AR 2251, §2.2.3).

EURAMET’s multimeter guide describes three sequences, and a procedure should say which one applies:

Source: EURAMET cg-15, §3.3.1–3.3.4. For sequence c), cg-15 gives an example of limits for skipping the adjustment on multimeters: 70 % of the annual specification at all points for less accurate instruments, 50 % for more accurate ones (§3.3.4). It is an example for one instrument class, not a general rule; set your own threshold and write it in the procedure. The adjustment itself follows the manufacturer’s method (§3.3.5). If repair or adjustment is needed just to make the instrument calibratable, agree it with the owner and document the parameters before and after (DKD-R 6-1, §5).

The sequence in the example procedure, as a decision flow:

1. 1

   Preliminary checks

   Inspection and functional tests.

   Fails: tag out, record as found "not calibratable"

2. 2

   As-found calibration

   Full sequence of test points, no adjustment of any kind. Record every reading.

3. 3

   Compare with the tolerance

   Apply the decision rule at every point.

   Any point out of tolerance: open an out-of-tolerance report

4. 4

   Decide whether to adjust

   Adjust only if a result passes the written threshold (0.50 psi in the example), using the manufacturer's method.

5. 5

   As-left calibration

   If adjusted, repeat the full sequence and record it as a separate set. If not, the as-found results are also the as-left results.

6. 6

   Record and label

   Results, uncertainty, standard, conditions, technician and date; status label; next due date.

Environmental conditions

ISO/IEC 17025 requires facilities and environmental conditions that don’t invalidate results, documented environmental requirements, and monitoring, control and recording where the method requires it or conditions influence the results, including on-site work (clauses 6.3.1–6.3.5, as quoted by PJLA). ISO 13485 asks for the same consideration in equipment procedures (clause 7.6). In a procedure, that means four things: the limits, the stabilization or warm-up time, what to record, and a rule that you don’t calibrate outside the limits. PJLA’s slides on clause 6.3 show exactly this kind of wording as an example.

Typical values from the sources, all as examples rather than universal requirements:

How tight to go depends on where the work sits in the traceability chain: the higher the standard, the tighter the control (NASA §10.2.1).

Measurement uncertainty in the procedure

ISO/IEC 17025 expects appropriate methods for evaluating measurement uncertainty where relevant (clause 7.2.1.1). NIST goes further: a calibration is not complete until the expanded uncertainty has been determined and reported, and every NIST SOP includes how to calculate it (NISTIR 6969, GLP 9). Published procedures from NIST, EURAMET and DKD all carry their own uncertainty section.

In practice, the procedure should name the uncertainty budget it uses (or include it), list the contributions to evaluate, and state a target uncertainty: the upper limit set by how the results will be used (VIM 2.6 Note 2; 2.34). Tie it to the decision rule, as in the example. Our measurement uncertainty guide explains budgets, and the uncertainty calculator builds one.

Method verification and validation (ISO/IEC 17025 clause 7.2)

Clause 7.2 is the part of ISO/IEC 17025 that most calibration procedure guides skip. It separates three situations:

Source: ISO/IEC 17025:2017 clauses 7.2.1.3–7.2.2.4, as quoted in PJLA’s slides on clause 7.2.

Techniques for validation include calibration or evaluation of bias and precision with reference standards or materials, comparison with other validated methods, interlaboratory comparisons, systematic assessment of influence factors, robustness tests and uncertainty evaluation (note 2 to clause 7.2.2.1). PJLA suggests the same kinds of techniques for verification. The performance characteristics to check can include range, accuracy, uncertainty, linearity, repeatability, reproducibility, robustness and bias, as far as they matter to the customer (7.2.2.3). If a method doesn’t specify environmental conditions that influence the results, document them in your procedure anyway: that follows from clause 6.3.

Three more points from clause 7.2:

• Manufacturer’s methods are acceptable. The standard lists methods specified by the equipment manufacturer among the recommended ones, next to those in international, regional or national standards (7.2.1.4). They still need verification.
• Deviations from a method are allowed only if documented, technically justified, authorized and accepted by the customer (7.2.1.7).
• Third-party procedures need checking. NASA notes that ASTM and ISA procedures are peer-reviewed by standards committees, while procedures posted on the NIST and GIDEP websites don’t go through an approval process. Its advice: treat every third-party procedure as guidance that needs verification and validation before use (NASA §10.2.1, p. 81).

Review, approval and document control

Procedures fail audits less often for what they say than for how they are controlled. EPA’s guidance on SOPs covers the essentials (EPA QA/G-6, §1.4 and §2–3):

• Written by people who do the work, in active voice and present tense, step by step and without ambiguity. Flowcharts help.
• Tried out by someone other than the author before approval, then reviewed and approved by qualified people.
• Controlled: document number, revision and date on every page, signatures (electronic ones are accepted), obsolete versions withdrawn and archived.
• Reviewed periodically, and updated whenever the process changes. EPA gives one to two years as an example of a review period, not a requirement. ISO 9001 and ISO/IEC 17025 set no frequency.
• Checklists support the procedure; they don’t replace it.

Under ISO/IEC 17025, methods and the documents that support them must be kept current and available to staff (clause 7.2.1.2), and a revised method needs its verification repeated as far as necessary (7.2.1.5). A good trigger for review is any change of reference standard, equipment model, referenced standard edition or tolerance, and every out-of-tolerance investigation that points at the method.

Calibration procedure checklist (printable template)

Use this list to draft a new procedure or to review an existing one before an audit. Each line maps to a section of the structure above. Print it, or copy it into your own template. It is free, with no sign-up.

Where to find existing calibration procedures

You rarely need to start from a blank page. These sources are free unless noted, and all of them are starting points that you verify and adapt, not ready-to-use procedures for your lab:

For instrument-specific procedures, see our instrument guides as they are published, starting with pressure gauges, scales and thermometers.

Common mistakes and myths

Related guides

• How to set up a calibration program: where procedures fit in the whole program.
• How to read and review a calibration certificate: what an external lab’s procedure should produce.
• Calibration records: the form your procedure fills in.
• Out of tolerance: what to do when a calibration fails: what the as-found data triggers.
• Calibration intervals: how often to run the procedure.
• Guard banding, the TUR calculator and the uncertainty calculator: the numbers behind your acceptance criteria.