Standardization

There are a number of standardization organizations, whether international (ISO, CEN, AACC, ICC, etc.) or national (Afnor, GOST, IRAM, GT, etc.), but what actually are standards?

These are documents which are voluntarily applied (or sometimes included in regulations) and have two main objectives:

• To describe the operating procedures of an analytical method.
• To provide quantified information for determining the performance achievable by the method when the protocol is followed.

What are standards used for? Mainly for establishing sound and agreed-upon rules for business transactions.

Imagine a world without standards: a buyer looking for wheat with 12% protein (and being aware that the analysis may vary) may require a seller to supply a batch of wheat with a protein level between 11.99% and 12.01%... a requirement that's difficult to achieve and which may result in an inability to deliver the goods or to a decrease in the price paid for not being within the specifications of the contract. On the other hand, the seller might only guarantee this batch to be between 9 and 15%... which could very well cause problems in production. They must therefore agree upon a value that is narrow enough to allow good batch separation which is not too high, in order to avoid negatively affecting business.

To set these values, the standards organizations rely on the work of independent experts who follow this basic process:

• Proposal of a new method (a draft of the future standard).
• Implementation of an inter-laboratory study. This is one of the key steps in the process. A dozen samples which are representative of what we are looking to analyze (in our case: protein levels) are sent to a dozen laboratories.
  • The choice of samples is important because it must cover the range of possible values (the standard is only valid over the range studied). In our case, a set of samples ranging from 11 to 13% would exclude from the scope of the standard anything below 11 or above 13.
  • If the standard is to be applied to other types of grains, they must be present in the set to be studied; we cannot extrapolate to corn what we measure for wheat. Hence, a study on flour cannot be extrapolated to wheat. There are currently methods which some people use for wheat when they have only been studied for (and therefore can only apply to) flour. [1] The standards therefore indicate a "field of application" which clearly specifies the conditions (i.e. types of samples) under which they can be used (e.g. flour, wheat, ground wheat, etc.). Anything not specifically mentioned in this "field" cannot be included in the standard.
  • In the field studied, the samples must be equally distributed. Five samples with 8% protein and five with 17% must be avoided.
  • The samples must correspond to what users of the future standard will test. For example, if grinding is necessary, this should be part of the test and participants should receive grain and not ground wheat.
  • The samples must be prepared and distributed to the participants. Users will receive packages with no other indication than a bag number.

• Laboratories must conduct the analyses within a given time period by scrupulously following the method. They will note any difficulties encountered and return the results to the method's organizer.
• Once all the results have been received, a statistical analysis will quantify the method's performance. Here are some indicators:
  • The repeatability standard deviation which measures the average deviation of the results obtained when the test is repeated by the same laboratory on the same device and with the same operator on the same day.
  • The reproducibility standard deviation which measures the average deviation of the results obtained when the test is performed by different laboratories, on different devices, with different operators on different days.
  • Logically, the standard deviation for reproducibility is always higher than the standard deviation for repeatability.
  • This base will be used to calculate other indicators such as repeatability or reproducibility limits, coefficients of variation, uncertainties, critical differences, etc. All this useful information is derived from the results of the inter-laboratory study.
  • A final important piece of information is whether the observed standard deviations are consistent regardless of the measured value, or whether they decrease or increase with the mean value (which means that we are respectively more or less accurate on the strong values).
• Experts then meet to validate these results and refine the wording of the standard based on the observations received from the laboratories in order to optimize its effectiveness.

We'd like to underscore the fact that a standard, without this numerical data, is no more useful to manufacturers than a simple user's manual. If you use a standard, it's mainly so you can apply this consensual data. Whether for your internal use (control charts) or for your business contracts.

What is also important to remember is that these standards are established FOR you, but must, above all, be established BY you. We encourage as many users as possible to get closer to their standards organizations in order to have an opportunity to participate in the establishment of these important documents. You can become participants or take part in discussions and/or inter-laboratory studies. It is both useful and rewarding. So why not take part in establishing the next standards document?

[1] This is very logical. Let's take the example of measuring water content. If performed directly on flour, the method will have a certain level of precision. If we receive wheat, we'll need to grind it. We understand that this step will add more room for error and uncertainty. The accuracy of the method stemming from wheat (more complicated) will not be as good as the method stemming from the analysis of flour (simpler). Therefore, any method applicable to wheat must be the subject of a specific study incorporating at least grinding or laboratory milling. This can be the subject of a specific standard (as in the case of the Alveograph) or integrated into a common standard (as in the case of the Mixolab 2), but in all cases this matrix must have been studied during the inter-laboratory study.