Rounding to Decimal Places: Rules, Examples and Pitfalls

Student marking a decimal place on worksheet

To round a number to a given number of decimal places, find that place, look at the single digit right after it, and raise the place digit by 1 if that next digit is 5 or more. If it is 4 or less, leave the place digit alone. Then drop everything to the right. That is the whole rule, and the rest of this guide shows how to apply it cleanly, where it goes wrong, and how to keep rounding from damaging a longer calculation.

TL;DR

Count decimal places from the decimal point: tenths is 1 place, hundredths is 2, thousandths is 3.
Look only at the digit immediately to the right of your target place. 5 or more rounds up, 4 or less stays.
Carrying happens: 0.99999 to 3 places is 1.000, and the trailing zeros stay because they show the precision.
In multi-step work, keep full precision and round once, at the end.
"5 rounds up" is a convention. Spreadsheets, code and statistics sometimes use other tie rules, so check which one your tool uses.

What "decimal places" actually counts
The four-step method
Rounding examples, side by side
Decimal places are not significant figures
Rounding and what the number is supposed to mean
When to round: once, at the end
More digits do not mean more accuracy
Ties, software and the 2.675 problem
Perspective: how we think about rounding at MetriqaHub
Check your rounding with our tools
Common mistakes and how to avoid them
Quick practice set
FAQ
Sources

What "decimal places" actually counts

Decimal places are counted from the decimal point, moving right. According to BBC Bitesize, 5.743 has 3 decimal places, 5.1492 has four, and 4.34 has two.

The names map to counts like this:

Instruction	Decimal places	Digit you keep up to	Digit you inspect
Nearest tenth	1	tenths	hundredths
Nearest hundredth	2	hundredths	thousandths
Nearest thousandth	3	thousandths	ten-thousandths
Nearest whole number	0	ones	tenths

So "round to the nearest tenth" and "round to 1 decimal place" are the same instruction. So are "round to the nearest hundredth" and "round to 2 decimal places." If you only remember one thing from the table, make it the last column: the digit you inspect is always the one just past the place you are keeping.

The four-step method

OpenStax presents rounding a decimal as a four-step procedure. In plain words:

Locate the place you are rounding to and mark it.
Look at the digit immediately to its right.
Decide: if that digit is 5 or more, add 1 to the marked digit. If it is less than 5, leave the marked digit as it is.
Drop all the digits to the right of the marked place.

A classroom rhyme covers step 3: four or less let it rest, five or more raise the score. It works, but only if you apply it to the right digit. The most common error is looking at the last digit of the whole number instead of the one next to your target place.

Worked example: rounding to the nearest tenth

Round 7.846 to the nearest tenth.

The tenths digit is 8. That is the place you keep.

The next digit, the hundredths digit, is 4.

4 is below 5, so the 8 stays.

Drop the rest: 7.8.

Notice that the 6 at the end is irrelevant. It would only matter if you were rounding to hundredths. Rounding 7.846 to the nearest hundredth gives 7.85, because the thousandths digit is 6.

Now the case from the common question "What is 218.653 rounded to 1 decimal place?" The tenths digit is 6, the hundredths digit is 5, so the 6 goes up to 7. The answer is 218.7. Again the 3 at the end plays no part.

Worked example: rounding to 2 decimal places

To round to 2 places, keep the hundredths digit and examine the thousandths digit. OpenStax's own example turns 18.379 into 18.38, because the 9 is at least 5, while 18.374 would become 18.37 because 4 is below 5 (OpenStax).

Two more of the same kind:

3.14159 to 2 places: the thousandths digit is 1, so the answer is 3.14.

0.0449 to 2 places: the hundredths digit is 4, the thousandths digit is 4, so the answer is 0.04.

That second one surprises people. The 9 at the end does not push anything up. You never round in chains from the right, one digit at a time. Doing that would turn 0.0449 into 0.045, then 0.05, which is wrong. Look at one digit only.

Worked example: rounding to 3 decimal places and carrying

Round 0.99999 to three decimal places. The thousandths digit is 9 and the next digit is 9, so the 9 goes up by 1. That makes it 10, so you write 0 and carry 1 to the left. The carry runs through every 9 in a row until it reaches the ones place. The result is 1.000.

The trailing zeros are deliberate. When an instruction asks for decimal places, 1.000 communicates three decimal places while a bare 1 may not (BBC Bitesize). Keep them.

Carry cases to practice:

Number	Places	Result
9.996	2	10.00
12.995	2	13.00
0.0999	1	0.1
199.96	1	200.0

Rounding examples, side by side

One number, several instructions, makes the pattern visible. Take 46.28571:

Round to	Digit inspected	Result
Whole number (0 places)	2	46
1 place (tenth)	8	46.3
2 places (hundredth)	5	46.29
3 places (thousandth)	7	46.286
4 places	1	46.2857

Each row starts from the original number, not from the previous row. If you rounded 46.28571 to 46.29 first and then rounded again to 1 place, you would get 46.3 here, which happens to match. But with a number like 2.4449, rounding twice (2.445, then 2.45, then 2.5) gives 2.5, while the correct one-step answer to 1 place is 2.4. Always return to the original digits.

Decimal places are not significant figures

The two are easy to confuse. Decimal places count positions after the point. Significant figures count meaningful digits starting from the first nonzero digit. OpenStax and BBC Bitesize both treat them as separate ideas (BBC Bitesize).

Take 0.004567:

To 2 decimal places: 0.00

To 2 significant figures: 0.0046

The first answer throws away nearly all the information. If your result is very small or very large, ask whether the task wants decimal places or significant figures before you start.

Rounding and what the number is supposed to mean

Rounding is not neutral. The House of Commons Library notes that rounding reduces precision, and that rounded percentages may not add up to exactly 100%. Its 2026 example takes 75.21% and gives 75.2% to 1 decimal place and 75% to the nearest whole number.

You can see the sum problem with three equal shares. Each is 33.333...%. Rounded to 1 place, each is 33.3%, and the total reads 99.9%. Nothing is wrong. The displayed figures just carry less information than the underlying ones. If a report needs the shares to total 100, say that the figures are rounded rather than quietly adjusting one of them.

When to round: once, at the end

This is the rule that separates tidy classroom answers from reliable work. A calculation is a sequence of operations, and rounding an intermediate result throws away information that later steps need. NIST's guide to SI conversion factors describes conversion as a multi-step process and gives the example 10.1 feet × 0.3048 = 3.07848 m. The sensible workflow: carry extra digits through every step and round once, for presentation.

A quick demonstration of the damage, using that same conversion. Suppose you cut the factor down to 0.30 early. Then 10.1 × 0.30 = 3.03 m. Done properly, 3.07848 rounds to 3.08 m at 2 places. The shortcut is off by 0.05 m, which is five centimeters on a length of about three meters, and the error would grow as you chain further steps.

A practical routine:

Write down the final precision you need (for example, 2 decimal places).

Carry at least two or three more digits than that through every intermediate step.

Round the final result once.

State the precision so the reader knows how to take it.

The same habit applies to money. If you compute sales tax of 7.25% on $19.99, the exact tax is $1.449275. Rounded to cents that is $1.45, for a total of $21.44. When many line items are involved, whether you round each line or only the grand total can change the final cent, so follow the convention that your invoice, tax authority or employer specifies.

More digits do not mean more accuracy

Showing 12 decimals on a screen looks precise. It may not be. NIST's rounding guidance for laboratories makes the point that retained digits should neither sacrifice accuracy nor exaggerate it, and that calibration results should be reported with an appropriate uncertainty (NIST GLP 9). It adds that expanded uncertainty is generally reported at roughly a 95% confidence interval.

For an engineer or a lab student, that changes the question. Instead of "how many decimal places should I use?" ask "how many decimal places can my measurement support?" A ruler marked in millimeters does not justify a result to four decimal places of a meter, however many digits the calculator shows.

Ties, software and the 2.675 problem

The schoolbook rule sends a 5 up. Software does not always agree.

Different tie rules exist

The exact halfway case, such as 2.5 rounded to a whole number, is called a tie. Python's decimal module documentation lists several modes: half-even (ties go to the nearest even value), half-up (ties go away from zero), half-down, ceiling, floor and truncation. So "5 rounds up" is a convention, not a law of software.

Here is how the same ties behave:

Value	Half up	Half even
0.5	1	0
1.5	2	2
2.5	3	2
3.5	4	4

Why would anyone choose half-even? The House of Commons Library notes that conventional half-up rounding creates a small upward bias across large datasets, because exact halves are always assigned upward, and it advises documenting the tie-breaking method in statistical work rather than assuming it (House of Commons Library).

Negative numbers add one more wrinkle. With the common "away from zero" version, -2.5 rounds to -3. With a "toward positive infinity" version, it rounds to -2. If your values can go negative, check which one applies.

Why a computer can disagree with your hand

A displayed decimal may not equal the value the computer stores. Python's tracker discussion shows that 2.675 is stored as a binary floating-point number slightly below the exact decimal, so round(2.675, 2) can return 2.67, while Decimal("2.675") gives 2.68 under half-even rounding (Python issue 24827). The rule was not broken. The number being rounded was never exactly 2.675.

The takeaway for anything involving cents: parse input as a decimal value and choose the rounding policy explicitly. Do not leave it to a default.

Perspective: how we think about rounding at MetriqaHub

We build unit converters and finance calculators, so rounding is where most of the small errors people report actually begin. Our view comes down to three habits.

First, the calculation and the display are two different jobs. A converter should work at full available precision and round only when it shows you the answer. If you copy a rounded figure into the next calculation, you have started a second, lossy conversion without noticing.

Second, the number of places should match the use. A recipe in cups rarely needs more than one or two decimals. A loan payment needs cents. An engineering conversion needs as many digits as the measurement can honestly support, and no more.

Third, state the rule. When rounding could change an answer, whether through ties, carries or the percentages-do-not-sum problem, a good result says what it did. Readers who know the rule can check the answer. Readers who do not can at least trust that the number was not quietly bent.

— Nkosi

Check your rounding with our tools

When you want to confirm a hand calculation, run it through the free converters and calculators at MetriqaHub. They cover length, mass, temperature and volume, along with finance calculators for things like loan payments and percentage change. There is no registration, and results appear instantly, so you can compare your rounded answer against the full-precision one in seconds.

Common mistakes and how to avoid them

Inspecting the wrong digit. Always the one immediately right of your target place. In 218.653 to 1 place, that is the 5, not the 3.
Chain rounding. Do not round 2.4449 to 2.445, then 2.45, then 2.5. Go from the original digits once.
Dropping trailing zeros. 1.000 and 1 are not the same instruction. If the task asked for 3 places, show 3.
Rounding early in a multi-step problem. Keep extra digits and round at the end.
Confusing decimal places with significant figures. Check what the question asks for.
Assuming software uses half-up. Look up the tie rule and, for money, use decimal arithmetic.
Forcing percentages to total 100. Note that figures are rounded instead.

Quick practice set

Try these, then check below.

56.472 to the nearest tenth

0.8651 to 2 decimal places

99.95 to 1 decimal place

3.14159 to 3 decimal places

0.99999 to 3 decimal places

Answers: 1) 56.5 (the hundredths digit is 7). 2) 0.87 (the thousandths digit is 5). 3) 100.0 (the hundredths digit is 5, and the carry runs all the way up). 4) 3.142 (the next digit is 5). 5) 1.000.

FAQ

How do I round a number to the nearest tenth?

Keep one digit after the decimal point and look at the hundredths digit, the second one after the point. If it is 5 or more, increase the tenths digit by 1. Otherwise leave it. For example, 7.846 becomes 7.8 and 218.653 becomes 218.7.

What is 0.99999 rounded to three decimal places?

It is 1.000. The fourth decimal digit is 9, which pushes the third digit up, and the carry moves through the run of 9s to the ones place. Keep the zeros, since they show that the answer is given to three decimal places.

Does 5 always round up?

In school math, yes, it is the standard rule. In software and statistics, not always: half-even rounding sends exact ties to the nearest even digit, and other modes exist (Python decimal documentation). If a result matters, such as money or a published statistic, find out which rule the tool uses.

Should I round before or after a conversion?

After. Carry extra digits through every step of the calculation and round once at the end, as NIST advises for unit conversions (NIST SP 1038). Rounding the conversion factor or an intermediate value first can shift the final answer.

Rounding to Decimal Places: Rules, Examples and Pitfalls