Report 080 · Lab Science
Under every quantitative number a laboratory reports there is a physical object: a bottle, an ampoule, a wafer, whose value someone else assigned. NIST publishes the paperwork for those objects, and the paperwork is more candid than the phrase people quote from it. The value has its own uncertainty, the certification expires, and the document says in plain language that establishing traceability of your result is your job.
An instrument does not measure a concentration. It produces a response, a count rate, an absorbance, an intensity, and somebody converts that response into a concentration using a calibration built from a material whose value was assigned elsewhere. Every quantitative result in analytical chemistry is a comparison, and the thing being compared against is usually sitting in a fridge with a certificate stapled to the outside of the box.
I spend my working life in microwave spectroscopy, where the same structure holds. You are never reading a property off the universe. You are reading an instrument that has been anchored to something, and the quality of the anchor sets the ceiling on everything downstream. Which is why the most useful hour a working scientist can spend is reading, all the way through, one certificate of analysis for a standard they already use.
So let us read one. NIST Standard Reference Material 1643f, Trace Elements in Water, certificate issued 28 April 2025. It is about as ordinary as reference materials get: a unit is "approximately 250 mL of acidified water in a polyethylene bottle, which is sealed in an aluminized plastic bag to maintain stability," containing nitric acid at roughly 2 percent by volume. It exists, in NIST's words, "primarily for use in evaluating methods used in the determination of trace elements in fresh water." Water with known amounts of metals in it. Nothing exotic. Everything interesting is in the fine print.
The certificate defines the term rather than assuming you know it:
"A NIST certified value is a value for which NIST has the highest confidence in its accuracy in that all known or suspected sources of bias have been investigated or taken into account."
Read what that sentence promises and what it does not. It is a statement about diligence regarding bias, not a claim of perfection. Known or suspected sources have been investigated. Unknown and unsuspected ones, by definition, have not. That is the strongest honest claim a metrology institute can make, and NIST makes exactly it, no more.
How the number was produced is described too. The certified mass fractions "are consensus estimates that blend the results of the gravimetric preparation value and a value determined by either inductively coupled plasma mass spectrometry (ICP-MS) or inductively coupled plasma optical emission spectrometry (ICP-OES)." NIST weighed known masses of each element into a tank of acidified water, using solutions from its own SRM 3100 series of single-element standards, then measured the result independently, then combined the two. The certified value is not one measurement. It is an agreement between a preparation and a determination.
Every certified value on that certificate carries an expanded uncertainty:
"The expanded uncertainty for each certified value is calculated as U = kuc where k is the coverage factor for a 95 % confidence interval and uc is the combined standard uncertainty calculated through the application of the Monte Carlo method described in the ISO/JCGM Supplement 1."
Two details in that are worth pulling out.
First, what goes into uc. The certificate says it represents "the combined effect of uncertainty components associated with the gravimetric preparation, the ICP-MS or ICP-OES determination, method bias, and stability." Stability is in there. The number accounts for the material changing over time in its bottle, which is a category of uncertainty most people never think to include in anything.
Second, the coverage factor is not 2. Nearly everyone who has taken a measurements course carries around k = 2 as the number you multiply by for roughly 95 percent coverage. In SRM 1643f's table, k is listed per element, and it varies: 1.9 for aluminum, 2.0 for arsenic, 2.1 for barium and beryllium, 2.2 for cadmium and boron. That is what falls out of a Monte Carlo propagation of the actual distributions rather than a normal approximation, and NIST prints the value it actually used for each analyte instead of rounding everyone to the familiar number.
Now look at the magnitudes. Arsenic is certified at 56.85 µg/kg with an expanded uncertainty of 0.37. Cadmium is 5.83 µg/kg plus or minus 0.13, which is 2.2 percent of the value. Boron is 150.8 plus or minus 6.6, about 4.4 percent. This is a national metrology institute's best characterization of a solution it prepared itself gravimetrically, and it still lands in the low percent for several elements.
That matters because of where those numbers go next. When you calibrate against this material, its uncertainty does not stay behind on the shelf. It becomes a component of the uncertainty of every result you compute from that calibration. It is a floor you inherit. A method reported with an uncertainty smaller than its calibrant's is not a better method, it is an incomplete uncertainty budget.
This is the line I would put on the wall of every laboratory that keeps standards past their usefulness:
"The certified values delivered by SRM 1643f are valid within the measurement uncertainty specified until 01 August 2027. The certified values are nullified if the material is stored or used improperly, damaged, contaminated, or otherwise modified."
Nullified is a strong word and it is the correct one. Certification is not a permanent property of the substance. It is a statement about a specific material, kept a specific way, for a specific period. Handle it badly and the bottle still contains what it contains, but the certificate no longer describes it, and you have a solution of unknown composition that looks authoritative.
The certificate's own revision history makes the point from another direction. It records: "28 April 2025 (Changed period of validity; editorial changes); 04 August 2023 (Changed period of validity, updated format, editorial changes); 08 August 2015 (Original certificate date)." The material was first certified in 2015 and its validity window has been reset twice since. NIST also commits to monitoring, and to issuing "an amended certificate through the NIST SRM website" if substantive technical changes affect the certification, then instructs users to verify they have the most recent version before making use of any value.
Which means the printed sheet in the binder is not the reference. The current version on NIST's site is the reference. A certificate photocopied in 2016 and filed carefully is a historical document.
There is a smaller condition hiding in the same certificate, and it is a good example of how specific these documents get. The uncertainty on the mass concentration values assumes "that the temperature at which the material will be measured is between 15 °C and 25 °C." The mass fractions are in µg/kg and the concentrations in µg/L, and converting between them runs through the measured density, 1.0101 g/mL with its own stated uncertainty. Certified values come with the conditions under which they are certified, and the conditions are load-bearing.
The word people actually use in procurement is traceable. NIST's own vocabulary publication, Special Publication 260-136, gives the formal definition, taken from the international vocabulary of metrology:
"Metrological traceability is now formally defined as the 'property of a measurement result whereby the result can be related to a reference through a documented unbroken chain of calibrations, each contributing to the measurement uncertainty' [VIM3:2.41]."
The clause that gets dropped when people paraphrase this is "each contributing to the measurement uncertainty." Traceability is not a certificate of correctness passed down a line. It is a documented sequence of comparisons in which every step adds a little more uncertainty than the step above it. Being traceable to the SI does not mean being accurate. It means being connected, with the cost of the connection accounted for.
SP 260-136 is explicit that the length of that chain has consequences. Where no suitable calibration material exists, traceability can be established by calibrating with a matrix material, but "the traceability chain for a matrix material has at least one more link than that of a calibration material and so will not provide the smallest possible measurement uncertainty." One more link, measurably more uncertainty. That is the whole concept in a sentence.
Here is the part that changes how you read a vendor's product page. NIST defines the shorthand and then assigns responsibility:
"'Traceable to NIST' is shorthand for 'Metrologically traceable to the SI (or other higher-order reference system) through measurement results certified by NIST.'"
And the policy that follows it, in three bullets: the provider of a measurement result "is responsible for supporting its claim of the traceability of that result or value," whether that provider is NIST or anyone else; the provider "must document the measurement process and uncertainty estimates used to establish the claim and provide a description of the chain of calibrations"; and the user of a result "is responsible for assessing the validity of a claim of traceability."
The figure caption in the same document says it about as directly as a federal publication ever says anything: NIST establishes that the certified values delivered by its SRMs are traceable when the material is properly stored and used, but "it is the customer's responsibility to establish the traceability of their measurement results by documenting that their measurement procedure is appropriately calibrated and validated."
So the phrase "NIST traceable," standing alone on a spec sheet or a quote, is not a property that transferred to the seller by purchasing a standard, and it does not transfer to your data by purchasing theirs. It is an assertion about a chain, and the document says out loud that whoever asserts it owes you the documentation, and that assessing it is on you.
One more distinction from the same publication, because it shows up on real certificates and is routinely skimmed past. Certificates carry certified values and also non-certified ones, and the two are held to different standards of claim. For non-certified values, "the documentation must not imply that NIST asserts that a non-certified value represents the true value of the measurand," and such values are printed in a separate section. Two numbers can sit on the same page in the same units and mean very different things.
This report is a close reading of two documents: the certificate of analysis for SRM 1643f and NIST Special Publication 260-136. Both were downloaded and read directly, and every quotation above is verbatim from one of them.
I have made no attempt to survey how laboratories actually handle expired or mishandled standards, how common out-of-date certificates are in practice, or how often uncertainty budgets omit the calibrant contribution. Those would be interesting empirical questions and I have no data on them, so I have deliberately not implied a prevalence anywhere in this piece.
SRM 1643f is one material chosen because it is common and unglamorous. Uncertainties, coverage factors and validity periods differ from material to material, so no number here should be read as typical of NIST reference materials generally. Read the certificate for the standard you actually use, not this one. Requirements from accreditation standards such as ISO/IEC 17025 also govern this territory in a working quality system, and I have not opened or cited them here.
Three things to carry out of a document that comes free in the box.
First, the standard has error bars. A certified value is NIST's best estimate with its known biases investigated, carrying an expanded uncertainty that can reach a few percent, computed with a coverage factor that is printed per analyte because it is not always 2. That uncertainty is inherited by everything you calibrate with it.
Second, certification is conditional and dated. It has an expiry, it is nullified by improper storage or use, it assumes a temperature range, and the certificate itself gets revised. The authoritative copy lives on NIST's site, not in your binder.
Third, traceability is a documented chain in which every link contributes uncertainty, and NIST's written policy puts the burden of supporting a traceability claim on whoever makes it and the burden of assessing it on whoever relies on it. When a supplier, a contract lab, or a dashboard says NIST traceable, the useful follow-up is not whether they have a certificate. It is which material, which certified value, from what date, and what the chain looks like between that bottle and this number. Where those questions have no answer, the phrase is decoration. Elsewhere on this site I have written about what a detection limit is really a property of, and this is the same lesson one level further down: a reported number is a claim about a measurement system, not a fact about the world.
U.S. Army combat veteran (Counter-IED / Electronic Warfare), peer-reviewed researcher in microwave spectroscopy, and founder & CEO of Shroombiosis. Consults on laboratory operations, AI, and supplement formulation.