First-party dataset

The US accredited calibration lab landscape, 2026

There is no single public register of accredited calibration laboratories in the United States. Several accreditation bodies operate here and each publishes its own directory in its own format. Each also has its own definition of what counts as one entry. This version covers five of them and checks every non-derived value against the page it came from. The counts are published separately here. Next to them sit a CSV carrying all 1,743 records, the limits of that coverage and the reasons a single national total would mislead.

Why there is no single number here

The obvious headline would be a national total. We are not publishing one because the data does not support it. We would rather say so than round the problem away.

Four things break the sum. A2LA lists accreditations per site rather than per company. A lab with three accredited locations appears three times, which is why 722 US rows resolve to 665 distinct organization names. ANAB does the opposite. It folds satellite sites into a parent record, so 43 of its 511 United States rows carry 171 further locations between them and its row count understates sites in the same way A2LA's overstates organizations. In its Forensic Calibration group ANAB does a third thing. It reuses one certificate across many sites. Four certificates cover 37 rows between them. The largest is a state breath-alcohol program and covers 21. Separately a laboratory may hold accreditation from more than one body. Adding the five directories together double counts anyone who appears in two of them. Name matching after normalization found eight organizations appearing in two directories. That is a floor rather than a measurement. It misses trading names, legal-name changes and branch naming, so the real overlap is higher by an amount this method cannot bound.

For completeness: the naive sum of US rows is 1,658 across A2LA, active PJLA, NVLAP, active ANAB and IAS. Deduplicating by normalized name gives 1,552. Neither belongs in a headline. Resolving it honestly needs address-level deduplication. That is the obvious next version of this study.

Where the labs are

State distribution below uses A2LA rows only. PJLA prints the city into the state column for 51 of its 760 worldwide records. Six of its 407 US rows carry a state that is not a two-letter code, three of them because the laboratory is outside the United States. The table keeps to A2LA rows. NVLAP does publish a usable state field, but at 78 US labs the distribution is too thin to mean anything: its largest state holds eight. A2LA US accreditations appear in 44 states plus Puerto Rico. ANAB's appear in 44 states plus the District of Columbia. IAS's appear in eight states. Two US rows carry no address at all, so the table below covers 720 of the 722 rows.

Top 12 of 44 states represented.

What they calibrate

PJLA and ANAB both publish calibration scope as a queryable category, with different vocabularies. The first table covers PJLA's 352 active US certificates. A laboratory usually holds several scopes so the column sums to more than 352.

ANAB uses ten categories of its own. This table covers the 422 active US certificates in its Calibration Labs group. 190 of them hold a single category and the average is 2.8, so here too the column sums to more than the number of laboratories. The 64 active certificates in its Forensic Calibration group all carry one category, Toxicology, and are not in this table.

Who the NVLAP labs are

NVLAP is the most institutional of the five. It is the only body that publishes a laboratory type. The split explains why. More than a third of its US calibration accreditations are government or state metrology laboratories rather than commercial providers.

When accreditations expire

Accreditation runs on a renewal cycle so expiration dates cluster. This is the shape of reassessment work through 2028 across the five directories. Twenty-six active ANAB certificates run to 2029 and five to 2030, and two IAS certificates run to 2030. Those 33 are not in the table. Two further active ANAB certificates carry a date written in a non-standard form. They are not in the table either and their expiration cell is empty in the CSV.

These are the expiration windows each directory currently exposes. Every A2LA US record falls inside 2026 to 2028 and NVLAP lists nothing beyond 2027, so the empty 2028 cell is a property of the source rather than a finding about renewal behaviour.

Using this to qualify a calibration supplier

A directory listing tells you a certificate exists. Some bodies also publish a status and some do not. NVLAP publishes none at all. So a listing on its own does not tell you that the accreditation is still live. It does not tell you whether it covers the work you are about to buy either. The scope document behind the listing tells you that, and every record in this study but one has one. The exception is named in Known limits. These are the checks the data says matter and the places a directory will mislead you if you read it quickly.

• The certificate is not the scope. Accreditation is granted for named measurements with stated uncertainties. A laboratory accredited for mass is not thereby accredited for torque. Four of the five bodies publish a scope document. IAS publishes a single certificate PDF per laboratory and lists no separate scope file, and that PDF is what the dataset links. The published dataset links it for 1,664 of the 1,743 rows. Nothing here has to be taken on trust.
• An accreditation is not always a calibration accreditation. 53 of the 722 A2LA US calibration records list further accreditations held by the same laboratory. Not one of those is a calibration accreditation. They are testing fields, reference material production, proficiency testing and inspection. A supplier quoting an accreditation number is not necessarily quoting the one that covers your instrument.
• Check at the body rather than from the supplier. A certificate that arrives by email tells you what was true on the day it was saved. The body's own directory tells you what is true now. For some laboratories it is also the only public record there is. 81 of the 722 A2LA US records carry no website at all.
• Check the date when you use the laboratory, not when you approve it. 303 of the accreditations counted on this page expire during 2026: 139 A2LA, 48 active PJLA, 30 NVLAP, 79 active ANAB and 7 IAS. Approving a supplier once and filing the certificate is how an expired accreditation ends up in your records.
• Read the status word carefully. 55 of the 407 US PJLA records in this dataset are Withdrawn or Suspended. 49 of those carry an expiration date that has already passed. PJLA also returns the same eight US certificates under both Active and Extended, because Extended means an active accreditation carrying an extension rather than a separate state. Directory data carries plain errors as well. One record here gives its country as the United States and its city as Dominican Republic.
• One directory is not the whole picture. A supplier can hold a real accreditation from a body you did not check, and two bodies can disagree about the same laboratory. Six laboratories in this dataset are listed as commercially available by one body and withdrawn, suspended or voluntarily withdrawn by another, at the same city. This version covers the five bodies we have found that accredit US calibration laboratories. If there is another, it is not here.

Every check above is a lookup that takes a few minutes. The reason it takes more than one lookup is that the United States has no single register, so there is no one place to ask.

Method and limits

Every record was collected from the accreditation body's own public directory. Each fetch was saved to disk with its HTTP status. For A2LA, PJLA, NVLAP and ANAB the manifest also records the source URL and the fetch timestamp. The seven IAS listing pages were saved with their response headers rather than a manifest. All 1,682 fetches behind this dataset returned 200. One earlier request to ANAB, refused by a bot rule, is not among them and is described under Known limits. Every non-derived field in all 2,451 collected records was then checked for verbatim presence in the specific page that record came from, and every check passed. Leading and trailing whitespace was stripped from every collected string. On the ANAB run that changed 95 values across 516 records, all of them a trimmed leading or trailing space. No interior character differs anywhere in the dataset. The check was itself tested against deliberately corrupted values before its result was trusted. On the three-body run 60 values were corrupted and all 60 were caught. On the separate ANAB run 6,841 values were corrupted in twenty shapes and all 6,841 were caught. On the IAS run 4,501 were corrupted and all 4,501 were caught.

Organization names were compared after lowercasing and removing every non-alphanumeric character. That choice moves the answer, so it is stated rather than hidden. Under stricter or looser normalization the A2LA organization figure ranges from 663 to 672 and the deduplicated cross-body total ranges from 1,547 to 1,560. The eight shared organizations are counted against active PJLA and ANAB certificates only. Counted against every status the figure is seventeen.

Two defects were found during that audit. Both are fixed in the published data. Two NVLAP laboratories that hold multiple accreditations had scope and certificate links pointing at the wrong program. Separately a collector treated a single-page result set as empty. That had dropped three suspended PJLA certificates. It would also have published a false zero for the ionizing radiation scope category, which actually has entries.

Known limits

• The CSV scope column is not comparable between bodies. In the published dataset the scope_categories field carries measurement disciplines for PJLA, such as Mass or Thermodynamic, but only the coarse directory programme grouping for ANAB, such as Calibration Labs or Forensic Calibration. The ten ANAB calibration sub-categories analysed on this page are a separate and finer field that the CSV does not carry. Comparing that one column across bodies will produce a wrong answer. We know because we did it to ourselves while checking this study and reached a confident conclusion that was false.
• Rows are not laboratories. A2LA counts per accredited site. Every row total on this page counts directory entries. Some figures count distinct organizations instead, such as the 665 A2LA organizations or the deduplicated 1,552. Those are labelled that way.
• Cross-body overlap is a floor, not a measurement. A lab holding two accreditations appears twice. Name matching found eight shared organizations, which is almost certainly an undercount.
• PJLA statuses overlap. Eight US certificates are returned by both the Active and the Extended query. Extended in PJLA means an active accreditation carrying an extension, not a separate state.
• Three PJLA records are misclassified at source. Each is labelled United States while sitting outside it. Yelfri Caribbean Calibration Services lists city and state as Dominican Republic. Datum Servicios Integrales de Medicion, certificate L24-91, is in Apaseo el Grande in Guanajuato. STC Labs is listed in Texas with a city that is in Mexico. The 407-row all-status US count carries all three. The 352-row active count carries two, because Datum is Withdrawn. We have left them as the directory publishes them rather than silently correcting them.
• Website coverage is incomplete. 81 of 722 A2LA US rows carry no website. That is the directory's gap, not ours.
• The bodies covered are the ones we found, not a guaranteed universe. This version covers A2LA, PJLA, NVLAP, ANAB and IAS. ANAB was held out of version one. Our first request to its directory was refused by a bot rule. A second and more ordinary client was served normally a minute later, and that run did return the directory. We set that copy aside unused and wrote to ANAB rather than publish data gathered either side of a block we did not understand. ANAB confirmed the directory is public. The collection published here was made after that, on 5 September 2026, and the earlier copy is still unused. IAS runs a separate Calibration Laboratories program that version one missed. It lists 133 accredited laboratories worldwide and 20 in the United States, and all 20 are now included. One further US body is known and not collected here. NAC is an ILAC Mutual Recognition Arrangement signatory for calibration and testing laboratory accreditation, and its directory lists two active US calibration laboratories. NAC belongs in the next version. If there are others we would like to hear about them. Bodies outside the United States are not covered either. A check of the UKAS directory on 5 September 2026, filtered to calibration laboratories in the United States, returned none, which is one reason to think that population is small. Both the IAS gap and the UKAS check were raised after publication by Christopher Paris of Oxebridge, and the study is better for it.
• NVLAP scope links are not in the dataset. NVLAP serves its scope and certificate documents through session-bound URLs that resolve to the directory search page once the session ends. Publishing one would hand you a link that looks live and is not, so the scope column is empty for all 78 NVLAP rows and the NVLAP lab code is in the certificate_number column instead.
• ANAB replaces the expiration date with status text on withdrawn rows. 25 of ANAB's 511 United States rows are not Active. On those the expiration field carries wording such as "Withdrawn as of 09/16/2025 until further notice" rather than a date, so the published CSV leaves expiration empty for them and no "already expired" count is possible for ANAB the way it is for PJLA.
• Two Canadian rows sit inside ANAB's United States query. ANAB's own country filter for the United States returned two laboratories in Ontario. They are in the dataset as Canadian rows, exactly as ANAB publishes them. They are not in the 486 or 511 United States figures on this page. Three Puerto Rico rows sit outside those two figures for the same reason, while the one A2LA Puerto Rico row is inside the 722. The country column carries three spellings for the same country, United States, USA and US, one per source, left as each directory publishes them. Filter on accreditation_body or on all three, not on one spelling.
• IAS publishes a certificate, not a separate scope document, and one of its links is empty. For IAS rows the scope column carries the certificate PDF, which is where IAS states the scope. One United States laboratory is listed with an empty certificate link. That is Cal-Cert Company and its cell is blank. IAS also prints city and state in one column, so the state field for IAS rows is derived by splitting it and is the only derived location field in the dataset.
• ANAB's forensic group reuses one certificate across many sites. ANAB accredits a separate Forensic Calibration group under ISO/IEC 17025. Its United States query returned 65 rows and 64 of them are Active. None overlaps the Calibration Labs group and all 65 are counted here. In that group one certificate can cover many locations. Four certificates cover 37 of the 65 rows. The largest is the Texas Department of Public Safety breath alcohol laboratory, which covers 21 sites on one certificate. The dataset carries every site as its own row, so within ANAB a certificate number is not unique. The scope column carries the group name for every ANAB row, so the two groups can be separated in the file.
• This is a snapshot. Directories change daily. Everything here reflects September 2026.

Common questions

How many ISO/IEC 17025 accredited calibration laboratories are there in the United States?

There is no single official count, because nobody publishes one. Each accreditation body maintains its own public directory and nothing joins them together. Counting the current directories covered here gives 1,658 active US accreditation records held by 1,560 organization names, which still double counts the eight organizations that appear in two directories, so the deduplicated figure on this page is 1,552. Neither figure is the number of laboratories. Records are accreditations rather than companies, a laboratory accredited by two bodies is counted twice, and some bodies fold satellite sites into a parent record. The honest answer is a range, and this study shows the per-body counts and the method behind them so you can see where each number comes from.

Which bodies accredit calibration laboratories in the United States?

The five bodies in this study are A2LA, ANAB, NVLAP, PJLA and IAS. They are not the whole universe. NAC, a United States body, is also an ILAC Mutual Recognition Arrangement signatory for the accreditation of calibration and testing laboratories, and it is not collected here. Accreditation from any ILAC signatory is recognized internationally through that arrangement, so a certificate from one is not more valid than a certificate from another. This study covers A2LA, PJLA, NVLAP, ANAB and IAS, taken from each body's own public directory.

Why can you not just add the per-body counts together?

Because the same laboratory can hold accreditation from more than one body, so it appears in more than one directory. Adding the columns double counts those laboratories. Three further problems push the same way: a directory row is an accreditation rather than a company, so one organization can hold several rows; at least one body folds satellite sites into a parent record, so one row can be many physical locations; and one body reuses a single certificate number across many sites. The study reports the naive sum, the name-deduplicated figure, and the gap between them rather than picking one and calling it the answer.

Where does this data come from and can I download it?

Every record was collected from the accreditation bodies' own public directories, the same pages anyone can search. Nothing is estimated, modelled, or bought. The full dataset is published as a CSV under a CC BY 4.0 licence, so you can check any figure on this page against the underlying rows or use it in your own work with attribution. The page also documents the defects found in the source directories rather than silently correcting them, because those defects change what the numbers mean.

Use the data

The dataset is free to use with attribution under a Creative Commons Attribution 4.0 licence. If you cite it please link to this page. Readers should see the method and the limits alongside the numbers.

Questions about the method, or a correction to a record? Email support@axiospec.com and we will look at the source page.