Significant Figures, Rounding and How a Result Is Reported

A lot comes back at 97.96% and the specification reads “not less than 98.0%”. Does it pass? The honest answer is that it depends entirely on a rounding convention, and that nothing about the material itself is in question. Arithmetic housekeeping decides the outcome, which is a good reason to understand the housekeeping.

Comparing a result with a limit

Convention says round the result to however many decimal places the limit carries, then compare. Under that rule 97.96 becomes 98.0 and the lot complies. The rule is defensible, and it also explains why a limit ought to be written at the precision its author actually meant.

Consider that “not less than 98%” and “not less than 98.0%” are two different specifications, not two spellings of one. Material sitting at 97.6% clears the first and fails the second. How a limit relates to what the method can actually do is covered in how a specification limit gets set.

Every digit is a claim

Writing a digit down asserts that the digit means something. Print 98.7% and you have claimed the tenths place is real. Print 98.70% and you have extended that claim to the hundredths. The longer number is not a more conscientious description of the material; it is a bolder assertion about the method.

What sets the number of defensible figures is the precision of the method, never the width of the instrument display. Software is perfectly willing to output 98.6741%, and four decimal places on a chromatographic purity is a rendering artifact rather than a measurement, since the spread described in measurement uncertainty typically overwhelms everything past that first decimal.

The same logic makes a trailing zero substantive rather than ornamental. Saying 98% and saying 98.0% are separate statements, because the zero declares that the tenths place was determined and came out zero, which is an assertion about resolution. Labeled quantities work identically: a vial marked 5 mg and a vial marked 5.00 mg differ in what they claim about how tightly the fill was controlled. This is the convention people manage to break in both directions at once, dropping zeros that were earned and adding zeros that were not.

Round once

Here is the rule responsible for most discrepancies that never needed to happen. Intermediate values keep extra digits, and rounding gets applied a single time, to the result that will actually be reported.

Round at every step instead and the error accumulates in whichever direction the arithmetic pushes it, which stops being trivial once three or four operations are chained together. Take a net content calculation that folds together chromatographic purity, water content and counter-ion content, of the kind laid out in net peptide content. That figure can shift by several tenths of a percent on rounding placement alone.

What to do with an exact five

When the digit being discarded is exactly five and nothing follows it, two conventions compete.

Neither convention is incorrect. What counts is choosing one ahead of time and sticking to it, since a laboratory that changes conventions between lots has manufactured a difference that will be read as a change in the material.

Four places where extra digits deceive

Chromatographic area percentage heads the list. Where the baseline gets drawn and where a shoulder gets split move the answer far more than any third decimal place could, as described in peak integration. Mass measurements come next, where the distinction is between calculated and observed. Four decimals on a calculated mass are meaningful; four decimals on an observed mass are meaningful only when instrument accuracy backs them up, which is the subject of mass accuracy in parts per million.

Third, anything sitting under the quantitation threshold. Report 0.03% using a method that quantifies no lower than 0.05% and you have written down a number the method cannot stand behind, per limit of detection and limit of quantitation. Fourth, unit conversions, where the trap is subtle. Turning 5 mg into 5000 µg is exact. Turning a measured 5.0 mg into 5000.0 µg conjures a digit out of nothing.

When two certificates differ in the last digit

Suppose one document says 98.7% and another says 98.65% for the same lot. Nothing happened to the material, and neither laboratory blundered. They simply made different reporting choices, following conventions that the paperwork almost never spells out.

So before drawing any conclusion, check what precision each result claims. A gap confined to a digit that neither method can resolve is not a disagreement about the substance at all, only two laboratories displaying numbers differently.

A gap in a digit both methods are capable of resolving is something else, and a genuine finding. Different columns, different gradients and different integration all drive that, and they are taken up in why certificates disagree on purity. Separating the two situations takes one look at how many figures are present, and it distinguishes a real discrepancy from a quarrel about rounding.