Can Something Be Close Mathematically but Far Apart Meaningfully?

Yes. Two numbers can be extremely close mathematically and still belong on opposite sides of a meaningful boundary. The size of the difference tells you how far apart the numbers are. It does not tell you how much the difference matters.

That distinction appears everywhere: a temperature just above freezing, a runner just outside a qualifying time, a product measurement barely beyond tolerance, or a probability just above a decision threshold. In each case, the numerical gap can be tiny while the consequences are discrete.

The important question is therefore not only “How far apart are these values?” It is also “What changes between them?”

Distance is not the same as consequence

Suppose a hypothetical manufacturing specification says a component can be no longer than 10.00 millimeters.

One component measures 9.99 mm. Another measures 10.01 mm.

Mathematically, they differ by only 0.02 mm. Under the hypothetical rule, however, one passes and the other fails.

Nothing mysterious happened to the arithmetic. We added something else: a decision rule.

That distinction matters because numbers do not carry their consequences around with them. People and systems establish tolerances, categories, goals, safety margins, eligibility requirements, and other boundaries. NIST describes this explicitly in conformity assessment: measurement uncertainty is a technical issue, while choosing a decision rule can involve the consequences and costs of accepting or rejecting something. (nist.gov)

So a small numerical difference can produce a large practical difference whenever it crosses a consequential boundary.

Sometimes the boundary is artificial

That creates a second problem. A sharp decision boundary can make the world look more sharply divided than it really is.

Imagine two students receiving hypothetical scores of 79.9 and 80.1 when 80 is the cutoff for a particular distinction. The classification changes at 80, but those scores alone do not establish that the students’ underlying knowledge suddenly becomes meaningfully different at that exact point.

The American Statistical Association has warned about an analogous problem in statistical analysis. It says scientific conclusions and business or policy decisions should not rest solely on whether a p-value crosses a particular threshold. Statistical significance also does not tell us the size or importance of an effect. (amstat.org)

That gives us two different ways mathematical closeness and meaningful distance can diverge:

A tiny difference can cross a boundary where the consequences genuinely change. Or a tiny difference can cross a boundary that merely sorts nearly identical cases into different categories.

The harder task is determining whether the boundary captures a meaningful change or creates the appearance of one.

Measurement itself has limits

There is another complication: the numbers can look more precise than the measurement supports.

Suppose a measurement is reported as 9.99 rather than 10.01. Before treating that 0.02 difference as decisive, you need to know something about how the measurements were obtained.

NIST notes that a measurement result is an estimate of the quantity being measured and is complete only when accompanied by information about its uncertainty. (nist.gov) A display containing more decimal places does not establish that an instrument can reliably distinguish differences that small. NIST specifically cautions that the number of displayed digits does not establish an instrument’s resolution. (itl.nist.gov)

That changes the question again.

Instead of asking only whether 9.99 and 10.01 are different, ask whether the measurement process can reliably distinguish them.

A mathematically visible difference can be practically important, practically irrelevant, or too uncertain to interpret confidently.

“Close” depends on what distance you measure

There is an even deeper version of the problem.

Consider two hypothetical businesses with annual revenue of $10 million and $10.1 million. If you are comparing company size by revenue, they are close.

Now imagine that the first company spent $9 million to generate its revenue while the second spent $15 million. The revenue figures remain close, but the businesses look very different if the question is financial performance.

The problem is not bad mathematics. We simply measured distance along one dimension and tried to turn it into a judgment about something multidimensional.

This happens with salaries, test scores, economic statistics, health measurements, rankings, AI similarity scores, performance metrics, and countless other quantities. A metric can answer the question it was designed to answer while failing to answer the question we actually care about.

The QuestionClass index contains a closely related question, What Does Visualizing Data Do for Decision Making?, which approaches the broader problem of turning numerical information into decisions.

Ask what happens in the gap

When two numbers appear close, a useful test is to investigate the space between them.

What boundary lies between the values? Who created it? What happens when it is crossed? How certain are the measurements? Would the difference still matter if the threshold moved slightly? And are you measuring the dimension that actually matters?

Those questions prevent two opposite mistakes.

The first is dismissing an important difference because “the numbers are basically the same.”

The second is exaggerating a trivial difference because the numbers happen to fall into different categories.

Mathematics can tell us that 9.99 and 10.01 are 0.02 apart. It cannot, by itself, tell us whether that gap means almost nothing or changes everything.

That requires context.

📚 Bookmarked for You

The Signal and the Noise by Nate Silver: A wide-ranging exploration of how to distinguish useful information from noise and reason more carefully about probability and uncertainty when numbers inform real-world judgments.

🧬 QuestionStrings to Practice: From Difference to Consequence

Use this sequence when a small numerical difference seems to be driving a surprisingly large conclusion.

How far apart are the values? → What boundary or decision lies between them? → How reliable is that boundary given the measurement uncertainty? → What actually changes when the boundary is crossed? → Would I make the same decision if the values moved slightly?

The sequence turns a comparison of numbers into an examination of consequences.

For more questions that help separate what we can measure from what we actually mean, visit QuestionClass’s Question-a-Day.

← Home