Exact, Approximate, and Measured Values

The Big Idea

Numbers can look equally precise on a page while making very different claims. The value \(12\) might be an exact count of twelve chairs, an estimate of twelve thousand customers, or a measured length rounded to twelve centimeters.

An exact value is known without approximation. An approximate value is deliberately close to another value. A measured value comes from observing a physical quantity and therefore has limited resolution and uncertainty.

The distinction matters because calculations cannot manufacture information that the inputs never contained. Good mathematical communication reports enough digits to be useful, but not so many that the result pretends to be more certain than the evidence allows.

Three Kinds of Numerical Claims

Every measured value is approximate, but not every approximate value is measured. The approximation \(3.14\) for \(\pi\) comes from rounding a mathematical constant, not from placing \(\pi\) beside a ruler.

Exact Values

An exact value names a quantity without rounding or measurement uncertainty. Common sources include:

• Counts: exactly \(28\) students are present, assuming the count and category are correct.
• Definitions: \(1\text{ m}=100\text{ cm}\) exactly.
• Algebraic results: the solution of \(3x=1\) is exactly \(x=\frac13\).
• Geometric formulas: a circle with exact radius \(r\) has exact circumference \(2\pi r\).
• Defined constants and structures: a square has exactly \(4\) sides.

Exact does not mean “written with only a few digits.” The values

\[ \frac13,\qquad\sqrt2,\qquad\pi,\qquad e \]

are exact even though their ordinary decimal expansions do not terminate.

Exact Forms Preserve Structure

An exact expression often communicates more than a decimal approximation. Compare

\[ \frac{\sqrt2}{2} \qquad\text{and}\qquad 0.7071. \]

The first expression identifies the value exactly and preserves its relationship to \(\sqrt2\). The second is convenient for calculation or reporting but has been rounded.

Fractions, radicals, powers, logarithms, trigonometric values, and symbols such as \(\pi\) frequently serve as exact forms. Keep them through symbolic work unless an approximation is needed for a practical purpose.

Approximate Values

An approximate value is close to an intended value but is not claimed to be equal to it. Approximations arise when we:

• round a long decimal,
• truncate digits,
• estimate mentally,
• use a numerical algorithm,
• replace a complicated model with a simpler one, or
• report a measured quantity with limited precision.

For example,

\[ \sqrt{20}\approx4.472, \qquad \frac13\approx0.333, \qquad 398\times51\approx20{,}000. \]

These approximations serve different purposes: numerical evaluation, finite decimal display, and estimation.

Use Notation That Matches the Claim

The symbol \(\pm\) does not explain how the uncertainty was obtained or what confidence level it represents. A complete technical report supplies that context.

A Finite Decimal Can Be Exact or Approximate

The written form alone does not determine the claim. As a mathematical number, \(0.333\) is exactly

\[ 0.333=\frac{333}{1{,}000}. \]

If it is being used to stand in for \(\frac13\), then

\[ \frac13\approx0.333. \]

If \(0.333\text{ m}\) is a measurement, it is a reported measured value whose final digit reflects the procedure’s precision. Context and notation tell us how the decimal should be interpreted.

Measured Values

Measurement compares a physical quantity with a unit through an instrument or defined procedure. A ruler, scale, clock, thermometer, sensor, survey, or laboratory method does not reveal infinitely many exact digits.

A reported measurement should identify:

• the numerical value,
• the unit,
• an appropriate number of digits,
• the uncertainty or resolution when relevant, and
• the method or conditions when they materially affect the result.

A statement such as \(24.6\) is incomplete if the reader needs to know whether it means meters, grams, seconds, degrees, dollars, or something else.

Counting and Measuring Are Different

A properly defined count is discrete and can be exact. If a class contains \(31\) enrolled students, the count is an integer rather than a rounded observation such as \(31.2\).

A measurement is continuous in principle and limited in practice. A table reported as \(180\text{ cm}\) long may be slightly longer or shorter than \(180\text{ cm}\), depending on the instrument, method, and rounding convention.

Counts can still be wrong because objects were missed, duplicated, or classified incorrectly. That is a data-quality problem, not inherent numerical measurement uncertainty.

Resolution

Resolution is the smallest change an instrument or reporting system can distinguish or display. A ruler marked every millimeter has finer resolution than one marked only every centimeter.

Resolution limits the detail available, but it is not automatically the entire uncertainty. Calibration, alignment, reading angle, environmental conditions, sampling, and the behavior of the measured object may add further uncertainty.

More displayed digits do not guarantee finer real resolution. A digital instrument may show several unstable or poorly calibrated digits.

Accuracy and Precision

Measurements can be precise but inaccurate if they cluster tightly around the wrong value. They can be accurate on average but individually imprecise if they are widely scattered around the correct value.

Error and Uncertainty Are Not the Same

Error is the difference between a result and a true or accepted reference value. Uncertainty describes the range or degree of doubt associated with a result.

If \(m\) is a measured value and \(t\) is an accepted reference value, then the signed error is

\[ m-t, \]

while the absolute error is

\[ |m-t|. \]

The true value may be unknown, which means the actual error may also be unknown. We can still estimate and report uncertainty from calibration records, repeated trials, instrument limits, sampling methods, or a validated model.

Relative and Percent Error

Absolute error must be interpreted relative to the size of the quantity. An error of \(1\text{ cm}\) is large when measuring a paperclip but small when measuring a highway.

For a nonzero reference value \(t\), relative error is

\[ \text{relative error}=\frac{|m-t|}{|t|}. \]

Percent error is

\[ \text{percent error} =\frac{|m-t|}{|t|}\times100\%. \]

If a measured value is \(49\) and the accepted value is \(50\), the absolute error is \(1\), the relative error is \(0.02\), and the percent error is \(2\%\).

Reporting Uncertainty

A symmetric uncertainty may be written as

\[ x=(12.4\pm0.1)\text{ cm}. \]

In a simple interval interpretation, this corresponds to

\[ 12.3\text{ cm}\le x\le12.5\text{ cm}. \]

Not all uncertainty is symmetric. A result may instead be reported as

\[ x=12.4^{+0.2}_{-0.1}\text{ cm}. \]

In statistics and experimental science, an uncertainty interval may carry a stated confidence or coverage level. The interval and its interpretation should be reported together.

Rounding Can Imply an Interval

If \(12.4\text{ cm}\) is known to have been rounded to the nearest tenth of a centimeter, then the unrounded value lies in

\[ 12.35\text{ cm}\le x<12.45\text{ cm}, \]

under the usual half-up boundary convention.

This interval follows from the rounding rule, not necessarily from the full physical uncertainty. The actual measurement process may justify a wider or differently shaped uncertainty range.

Significant Figures

Significant figures are digits treated as meaningful in a reported value. Common introductory conventions include:

• Every nonzero digit is significant.
• Zeros between significant digits are significant.
• Leading zeros are placeholders and are not significant.
• Trailing zeros to the right of a decimal point are normally significant in measured data.
• Trailing zeros in a whole number without a decimal marker may be ambiguous.

Decimal Places and Significant Figures

Decimal places count positions to the right of the decimal point. Significant figures count meaningful digits beginning with the first nonzero digit.

The value \(0.00450\) has five decimal places but three significant figures. Scientific notation makes this especially clear:

\[ 0.00450=4.50\times10^{-3}. \]

Decimal places are tied to place value. Significant figures are tied to the reported precision of the quantity.

Introductory Reporting Rules for Calculations

When uncertainty values are unavailable, classrooms and laboratories often use significant-figure rules as practical reporting conventions:

• Addition and subtraction: round to the least precise decimal place among the measured inputs.
• Multiplication and division: report no more significant figures than the measured input with the fewest significant figures.
• Exact counts and defined conversion factors: do not limit the result’s significant figures.

These are communication rules, not a complete theory of uncertainty. When actual uncertainties are available, propagating those uncertainties is more informative.

Keep Guard Digits Until the End

Rounding every intermediate result can compound error. Keep the exact form or several extra guard digits during the calculation, then round the final result once.

For example, if a calculation uses \(\sqrt2\), retain \(\sqrt2\) or a long calculator value internally rather than replacing it immediately with \(1.4\). The final reporting precision should reflect the problem’s inputs and purpose.

Calculator Digits Are Not Automatic Evidence

Suppose a calculator uses a measured radius \(r=2.3\text{ cm}\) and displays

\[ \pi r^2=16.619025137\ldots\text{ cm}^2. \]

The display does not make all those digits meaningful. Under the introductory significant-figure convention, the input \(2.3\text{ cm}\) supports a result of about \(17\text{ cm}^2\). A fuller uncertainty analysis may justify a different reporting form.

A calculator carries digits; it does not decide which digits the evidence supports.

Uncertainty Travels Through Calculations

If inputs vary within ranges, the output usually varies too. For quantities reported as \(x\pm\Delta x\) and \(y\pm\Delta y\), a conservative bound for addition or subtraction is

\[ \Delta(x\pm y)\le\Delta x+\Delta y. \]

For products and quotients with small nonzero uncertainties, relative uncertainties are often approximated by combining the input relative uncertainties according to an appropriate propagation method.

The safest elementary method is often to calculate with the allowed endpoints and determine a resulting interval. More advanced work distinguishes worst-case bounds, statistical uncertainty, dependence among inputs, and confidence levels.

Exact and Measured Conversion Factors

Some unit conversions are exact because the units are defined that way:

\[ 1\text{ in}=2.54\text{ cm} \]

exactly. This conversion factor does not add uncertainty or limit significant figures.

Other factors come from measurement, experiment, market data, or modeling. A material density, exchange rate, population estimate, or experimentally determined constant may be approximate and time- or condition-dependent.

Before using a conversion factor, ask whether it is defined, measured, estimated, or rounded.

Tolerance Is a Requirement, Not a Measurement

A tolerance specifies the acceptable departure from a target. If a part is specified as

\[ 10.00\pm0.05\text{ mm}, \]

then values from \(9.95\text{ mm}\) through \(10.05\text{ mm}\) meet that stated dimensional tolerance.

Tolerance is not the same as measurement uncertainty. The tolerance describes what is acceptable; uncertainty describes how confidently we know the part’s actual dimension. A measurement system must be capable enough to decide whether the part meets the tolerance.

Model Values Are Approximate for Another Reason

A model may use exact arithmetic and still produce only an approximate description of reality. A population model might predict exactly \(P(10)=42{,}617.38\) within its equations, but the prediction depends on assumptions about growth rates, migration, data quality, and future conditions.

It helps to separate:

• computational error from numerical approximation,
• measurement uncertainty in input data, and
• model uncertainty from simplified assumptions.

A mathematically precise calculation is not automatically an exact forecast.

Bounds Can Be More Useful Than a Long Decimal

Sometimes the most responsible statement is not a central estimate but a guaranteed range. For example,

\[ 1.414<\sqrt2<1.415 \]

gives explicit lower and upper bounds. Bounds are especially useful in proof, safety planning, budgeting, engineering tolerances, interval arithmetic, and error control.

A one-sided bound may also be appropriate: a bridge must support at least a stated load, or a contaminant level must remain below a legal limit.

Choose Precision for the Decision

More digits are not always better. The useful level of precision depends on what the result will be used to decide.

• A grocery estimate may only need the nearest dollar.
• A medication dose may require much tighter control and professional standards.
• A construction cut may need millimeter precision.
• A national population may be responsibly reported to the nearest thousand.
• A proof should retain exact forms whenever approximation would weaken the conclusion.

Precision should be sufficient for the operational requirement and supported by the source data.

A Responsible Numerical Workflow

1. Identify the source. Was the value counted, defined, derived, measured, estimated, or modeled?
2. Classify the value. Is it exact, approximate, measured, or a combination?
3. Record units and conditions.
4. Preserve exact forms during symbolic work.
5. Retain guard digits during numerical work.
6. Track uncertainty, bounds, or significant figures.
7. Choose the final precision for the actual purpose.
8. Round once at the end.
9. Use \(=\), \(\approx\), \(\pm\), or interval notation correctly.
10. Check whether the result claims more certainty than the inputs justify.

Common Confusions

Working Summary

Exact values preserve the complete mathematical value. Approximate values trade some exactness for convenience. Measured values report observations with limited precision and uncertainty.

Accuracy concerns closeness to a true or accepted value. Precision concerns repeatability or reported detail. Resolution concerns the smallest detectable change. Error compares a result with a reference; uncertainty describes how much doubt surrounds the result.

Use notation honestly, retain exact forms and guard digits during work, round at the end, attach units, and never let a calculator display create false confidence.

Check Your Understanding