A safety data sheet with a column of "no data available" entries is compliant and unhelpful. Customers read those gaps as a quality signal, procurement departments ask about them, and each one is a question someone will raise again. Some of them can be closed with a calculated or estimated value, properly labelled; others cannot and should stay open rather than be filled with a number that looks measured. Knowing which is which takes an afternoon per substance and removes a recurring source of friction.
The gaps worth separating
Not all empty fields have the same cause, and the remedy differs:
- Never determined. No study exists because nobody needed the value. Often the easiest to close, sometimes with a calculation, sometimes with a cheap determination.
- Not applicable. The property has no meaning for this form, for example a flash point for an inorganic solid. The honest entry says not applicable with the reason, not "no data available".
- Technically not determinable. The substance decomposes before boiling, or the method cannot be applied. State that, with the reason, which is more informative than either a blank or a number.
- Determined but lost. The data exist in an old report nobody can find. Worth searching before any modelling, because a measured value with a method reference beats every estimate.
Classifying the gaps this way usually shows that a third of them are not data problems at all.
Where a calculated value is normal practice
For several physicochemical properties, estimation from structure is established and widely accepted in the relevant sections, as long as it is identified as calculated. The partition coefficient is the clearest case: a calculated log Kow with the method named is a routine entry, and for many substances no measured value exists. Other properties where a calculated or estimated entry is common include vapour pressure, water solubility as an order-of-magnitude indication, and dissociation constants.
Two conditions make these entries defensible. The method is named, so a reader can judge it, and the value is unambiguously marked as calculated rather than sitting in the same format as measured data. An estimate presented as a measurement is the one outcome worse than a gap.
Where a gap should stay a gap
Entries that feed a hazard classification, a threshold comparison or an exposure assessment should not be filled with an estimate. In practice that means the classification-relevant toxicological and ecological endpoints, physical hazard properties determined by a specified test such as flash point or oxidising properties, and anything a downstream user will use as an input to their own risk assessment. For these, the correct action is to state the gap, and separately to plan whether the study is needed for the tonnage band and markets involved.
The judgement about which entries are classification-relevant for a specific substance, and which adaptation routes apply, belongs with a qualified assessor. The structure above is for organising the work before that conversation, not for replacing it.
How to write a closed gap
A usable entry has four parts, and the fourth is the one usually missing:
| Part | Example of what it answers |
|---|---|
| The value with its unit and the conditions | At which temperature and pressure, for which form |
| The basis | Measured, calculated or read across from a named analogue |
| The method or model, with version | The test guideline, or the estimation method and its version |
| The date and the person or system responsible | So the entry can be reviewed and superseded |
Keeping this in the underlying substance record rather than only in the document means the next revision, the next language version and the next customer question are all answered from one source. It is also the difference between a sheet that can be updated and one that has to be reconstructed.
Treat the substance record as the product, not the document
Most of the recurring effort in safety data sheet maintenance comes from the same values being held in several places: one in the authoring system, one in a spreadsheet, one in the old report, with no record of which is current. A single substance record per property, with basis, method and date, generates every document and makes a gap analysis a query rather than a reading exercise. For a portfolio of a few hundred substances this is the change that reduces the work, not better templates.
A practical first pass: list every property field across the portfolio, count the gaps per field, and look at the top five. They are usually the same five fields for every substance, which turns several hundred individual problems into five decisions.
Limits of the statement
- This article does not tell you which entries your sheets require. The applicable annex, its current wording and the guidance for your markets govern, and they change.
- No calculated value closes a gap that the regulation requires to be determined by a specified method, whatever the model's quality.
- Mixtures follow their own rules, and a property of a mixture is frequently not derivable from the properties of its components.
- A classification, and the decision whether an adaptation route applies to a specific endpoint, requires a qualified assessor. Organising the data gaps is preparation for that work.
Frequently asked questions
Is "no data available" acceptable in a safety data sheet?
It is a permitted entry where no information exists, but it is worth distinguishing from "not applicable" and from "not technically determinable", each stated with its reason. Those two are more informative and often account for a large share of apparent gaps.
Can a calculated value be entered instead of a measured one?
For several physicochemical properties this is established practice, provided the value is clearly identified as calculated and the method is named. For classification-relevant endpoints and for properties the regulation ties to a specified test method, it is not.
What is the most common gap worth closing?
The partition coefficient, where a calculated log Kow with the method named is routine and a measured value frequently does not exist. In a portfolio review the same handful of fields usually accounts for most gaps.
How should an estimated entry be labelled?
With the basis stated as calculated, the method or model and its version named, and the entry visually distinct from measured data. An estimate that looks like a measurement is worse than an open gap.
What reduces the recurring maintenance effort?
Holding each property once in a substance record with its value, basis, method and date, and generating the documents from it. Most of the repeated work comes from the same value existing in several places with no record of which is current.