BTT Mathematics / Primary Mathematics Learning Hub
A check can expose a wrong answer without being able to certify a right one. Digit sums are a useful place to learn that distinction. They preserve a small piece of a number’s structure, making some arithmetic mistakes easy to detect. They also deliberately discard information, allowing other mistakes to pass unnoticed.
Consider 368 + 457. The correct total is 825. A proposed answer of 835 fails a remainder-nine check. A proposed answer of 852 passes the same check even though it is wrong. An answer of 915 passes both the remainder-nine check and the last-digit check. These outcomes are not contradictions. Each check is asking a limited question.
The aim of this guide is therefore not to teach a magic way to verify everything quickly. It is to show what the digit-sum method preserves, why it works, which errors it catches, and when another check is needed. Every final answer should still be connected to the original calculation and the original problem.
This is Primary Mathematics enrichment. Begin with place value, multiplication tables and remainders. The later language about modular checks and information loss is optional. No banking, identification, software-security or other real-world check-digit standard is being prescribed. The codes near the end are explicitly invented teaching examples.
Remainders and digit sums · Operation checks · Missed errors · Decimals · 24 questions · Worked solutions
1. Reduce a number without pretending to preserve all of it
Adding the digits of 462 gives 4 + 6 + 2 = 12. Adding again gives 1 + 2 = 3. This small number is connected to the original: both 462 and 12 leave remainder three when divided by nine.
However, many other numbers have the same remainder: 3, 12, 21, 30 and so on. Therefore remainder three does not identify 462. It identifies one class of numbers separated by multiples of nine.
Think of the check as a coarse label. Different numbers may carry the same label. If two quantities are truly equal, they must share that label. But sharing the label does not force them to be equal. This one-way implication is the foundation of the entire method.
Use remainders zero through eight
A remainder on division by nine is one of 0, 1, 2, 3, 4, 5, 6, 7 or 8. In this guide, reduce all checking values to that range. A multiple of nine receives checking value zero.
The term digital root usually means repeatedly adding digits until a single digit remains. For a positive multiple of nine, that single digit is nine, not zero. Thus 999 has digital root nine but remainder zero on division by nine. Both conventions can be used consistently, but mixing them creates avoidable confusion.
| Number | Repeated digit sum | Remainder on division by 9 |
|---|---|---|
| 0 | 0 | 0 |
| 18 | 9 | 0 |
| 462 | 3 | 3 |
| 999 | 9 | 0 |
| 10024 | 7 | 7 |
2. Why the place-value structure makes the rule work
Ten is nine plus one. One hundred is ninety-nine plus one. One thousand is nine hundred and ninety-nine plus one. Every whole-number decimal place value is one more than a multiple of nine.
Expand 462 as 4 × 100 + 6 × 10 + 2. Replace 100 by 99 + 1 and 10 by 9 + 1. The result is 4 × 99 + 6 × 9 + (4 + 6 + 2). The first two parts are multiples of nine. Removing them leaves the digit sum, twelve, without changing the remainder.
Reduce twelve again: twelve is nine plus three. Therefore 462 leaves remainder three. The difference 462 − 12 = 450 is divisible by nine, which gives another direct check of the argument.
The same reasoning applies to additional digits because every place contributes a multiple of nine plus its digit. This is a place-value proof, not a pattern inferred from a few examples. A related external explanation appears in the NRICH source listed at the end.
Why adding the digits works again
The first digit sum is itself an ordinary whole number, so the same rule applies to it. Repeating the operation does not change the remainder-nine class. Once the reduced result is small enough, the final remainder can be read directly.
For 572, the digit sum is fourteen. Fourteen leaves remainder five, so 572 also leaves remainder five. The method does not say 572 equals five. It says their difference is a multiple of nine.
3. Check addition by comparing remainder labels
When two whole numbers are added, their remainder-nine values can also be added and reduced again. Multiples of nine removed from the original inputs would contribute only more multiples of nine to the total.
For 368 + 457, the first number has digit sum seventeen and remainder eight. The second has digit sum sixteen and remainder seven. Add eight and seven to get fifteen, which leaves remainder six. Any correct total must therefore have checking value six.
The correct total, 825, has digit sum fifteen and checking value six. It passes. The candidate 835 has digit sum sixteen and checking value seven. It fails, so the written equality 368 + 457 = 835 is false.
What a mismatch tells us
If the check itself is calculated correctly, different remainders prove that the stated equality is wrong. They do not identify the cause. Perhaps an input was copied incorrectly, an addition was performed incorrectly, or the answer was mistyped. Return to the written quantities before deciding where to repair.
The test cannot evaluate the original story. A perfectly correct sum of the wrong two prices may pass every arithmetic check while still answering the wrong question. Model selection and arithmetic verification are separate responsibilities.
What a match does not tell us
The wrong total 852 also has digit sum fifteen and checking value six. It passes. The digits have been rearranged, but the check has no memory of their positions. The appropriate conclusion is “not rejected by this check”, not “proved correct”.
A last-digit check catches this particular error: eight plus seven ends in five, not two. But the wrong candidate 915 ends in five and also has checking value six. A size bound catches it: 368 is less than 400 and 457 less than 500, so their sum is less than 900. The candidate 915 is too large.
4. Check multiplication without losing the limitation
For multiplication, multiply the remainder-nine values and reduce the product. To see why, write each number as a multiple of nine plus its remainder. When the two expressions are multiplied, every term except the product of the two remainders contains a factor of nine.
Consider 247 × 36. The first factor has digit sum thirteen and checking value four. The second has digit sum nine and checking value zero. The product must have checking value zero.
A direct decomposition gives 247 × 30 = 7410 and 247 × 6 = 1482. Adding produces 8892. Its digits sum to twenty-seven, so its checking value is zero, as expected.
A near miss and a hidden error
The proposed product 8893 has digit sum twenty-eight and checking value one. It fails. The proposed product 8982 has digit sum twenty-seven and checking value zero. It passes, though it differs from the correct product by ninety.
The last-digit check also accepts 8982: seven times six ends in two. Even a reasonable broad bound can accept it. Because 240 < 247 < 250, multiplying by positive 36 gives 8640 < 247 × 36 < 9000. The wrong answer 8982 lies within that interval.
Several successful checks can therefore still leave an incorrect candidate. The exact decomposition 7410 + 1482 = 8892 resolves the calculation. Checks save effort when they expose errors; they are not a substitute for every exact calculation.
5. Check subtraction by rebuilding the minuend
For 803 − 476, a convenient Primary-level check avoids negative remainders. If the proposed difference is D, then 476 + D should recover 803. Compare the remainder labels of that addition.
The checking value of 803 is two. The checking value of 476 is eight. The correct difference 327 has checking value three. Eight plus three is eleven, whose remainder-nine value is two. Thus the required relationship is satisfied.
The wrong difference 337 has checking value four. Eight plus four reduces to three, not two, so it is rejected. The wrong difference 372 has checking value three and therefore passes the same check as 327.
Use an exact inverse check to settle it: 476 + 327 = 803, whereas 476 + 372 = 848. The remainder check is a compressed inverse relationship; the full inverse calculation retains more information.
6. Check division through multiplication and the remainder
For whole-number division with positive divisor, use the relationship:
Dividend = divisor × quotient + remainder, with 0 ≤ remainder < divisor.
The numerical identity and the remainder bound are both required. Do not attempt to divide the checking values directly. A divisor that is a multiple of nine has checking value zero, and division by that checking value would not make sense.
Worked example: a correct quotient and remainder
157 divided by 12 gives quotient thirteen and remainder one. Rebuild exactly: 12 × 13 + 1 = 156 + 1 = 157. The remainder one is between zero and eleven, so the normal division conditions are satisfied.
The compressed check also passes. The checking values of 12 and 13 are three and four. Multiply and add the remainder: 3 × 4 + 1 = 13, which reduces to four. The dividend 157 has digit sum thirteen and also reduces to four.
Two different ways an answer can be wrong
A proposed quotient sixteen and remainder one gives 12 × 16 + 1 = 193, not 157. Yet 193 and 157 share checking value four, so the remainder-nine filter misses the error.
A proposed quotient twelve and remainder thirteen satisfies the exact identity 12 × 12 + 13 = 157. However, thirteen is too large to be the remainder after division by twelve. Another full group of twelve can still be removed. The canonical result is quotient thirteen, remainder one. Exact reconstruction alone must be combined with the remainder bound.
7. Understand the errors this check cannot see
Two whole numbers have the same remainder-nine label exactly when their difference is divisible by nine. Therefore, for fixed correct inputs, any candidate answer differing from the true answer by a multiple of nine will pass this particular filter.
This is a precise limitation, not a vague warning that the method is “sometimes unreliable”. It tells us how to construct counterexamples and how to choose a different check.
Digit swaps preserve the sum
The numbers 348 and 834 have the same three digits, so their digit sums are both fifteen. Every rearrangement of a fixed list of digits preserves its sum. A remainder-nine check cannot detect a pure transposition or more general digit permutation on its own.
Leading zeros can change how a numeral is written, but they contribute nothing to its digit sum. The check still does not restore positional information. A written 0348 and 348 share the same value; 834 is a different value with the same checking label.
A single changed digit is usually detected, with one important exception
If one digit is replaced by another at a fixed position, the numerical change is the digit difference multiplied by a decimal place value. Since that place value has remainder one on division by nine, the change is invisible only when the digit difference is a multiple of nine.
For two different ordinary digits, that happens only with a change between zero and nine. For example, changing 102 to 192 replaces a zero with a nine and adds ninety. Both numbers have checking value three. By contrast, changing 462 to 472 adds ten, changing the checking value from three to four, so the substitution is detected.
This claim is limited to one fixed-position digit substitution with everything else unchanged. Two or more errors can compensate for one another. Do not use the single-error result as a guarantee for arbitrary corrupted calculations.
Appending a zero also escapes
Multiplying a whole number by ten appends a zero in decimal notation. Its digit sum remains unchanged. Thus a factor-of-ten error can pass the digit-sum check. That is why decimal placement and unit scaling need their own verification.
8. Put decimal quantities on a common whole-number scale
Digit sums in this guide apply to whole-number numerals. For a decimal addition, first rewrite all quantities using the same fractional unit. In 4.68 + 7.59 = 12.27, use hundredths: 468 + 759 = 1227.
The checking values are zero for 468, three for 759, and three for 1227. The scaled equality passes. The original exact addition is 12.27. But the raw digit-sum method would not by itself distinguish 12.27 from 1.227, because moving a decimal point does not change the digits.
Different decimal lengths still need equal units
For 4.68 + 7.5, rewrite the second number as 7.50 before scaling. The whole-number calculation is 468 + 750 = 1218, corresponding to 12.18. Using 468 + 75 would mix hundredths with tenths.
This is the same underlying issue as adding unlike measurement units. A correct operation on mismatched units is not a valid model of the original quantities. The Decimals and Four Operations guide develops that place-value foundation.
Decimal multiplication changes the scale twice
For 1.2 × 0.3, calculate twelve tenths times three tenths. The whole-number product is 12 × 3 = 36, but the unit is hundredths, giving 0.36. The result 3.6 has the same non-zero digits and cannot be rejected merely by comparing their digit sums.
A magnitude check is decisive here: multiplying positive 1.2 by a positive factor below one makes it smaller than 1.2. The candidate 3.6 violates that bound. Place value gives the exact result; the bound rejects the misplaced-point answer.
9. A toy check digit demonstrates information loss
Invent a classroom rule: append one digit from zero through eight so that the sum of all digits is divisible by nine. For payload 4826, the digit sum is twenty. Add a check digit seven to reach twenty-seven. The complete toy code is 48267.
If the code is later written as 48268, the sum becomes twenty-eight, so it fails the rule. But 84267 passes: the payload digits four and eight have been swapped, preserving the sum. The check detects some changes and misses others.
This toy rule does not authenticate a person, prove the accuracy of a record, or protect information from deliberate alteration. Anyone who knows the rule can construct a different code that satisfies it. Its educational value is showing what a simple consistency condition retains and what it throws away.
The check digit is selected, not guessed
Reduce the payload digit sum to remainder r. Choose the unique digit c from zero through eight that makes r + c a multiple of nine. If r is zero, choose zero. If r is two, choose seven. If r is six, choose three.
Restricting the allowed check digit to zero through eight avoids giving two choices, zero and nine, when the payload already has remainder zero. The convention is part of the invented rule.
10. Combine checks without overstating the conclusion
Use a check suited to the likely error. A last-digit check can catch a units error. A rough estimate can catch an extra zero. A remainder-nine check can expose many digit substitutions. An inverse operation can verify an exact numerical relationship. Returning to the question checks whether the relationship was the one required.
Different checks should not be treated as statistically independent votes. For example, divisibility by three and remainder nine are closely related. Passing both does not double the evidence in a meaningful numerical sense. We are checking logical conditions, not calculating a probability of correctness.
Sometimes a proven narrow range and a remainder condition together can isolate one answer. If an integer is known to be between 820 and 830 inclusive and to have remainder six on division by nine, only 825 fits. That conclusion uses the range as well as the remainder. It does not show that a matching digit sum alone certifies a general calculation.
The practical sequence is therefore: interpret, estimate, calculate, check a different feature, then return to the original conditions. Do not spend more time performing elaborate checks than a simpler exact recalculation would require.
11. Practice: 24 original questions
Use checking values from zero through eight. When a candidate passes, say only that the remainder-nine check does not reject it unless an exact calculation or complete argument also verifies it.
Questions 1–8: Remainders and addition
- Find the remainder-nine checking value of 462.
- Find both the digital root and the remainder-nine checking value of 999.
- Find the checking value of 10024.
- Find the checking value of 572.
- For 368 + 457, test candidate 825 using remainder nine.
- Test candidate 835 for the same addition.
- Test candidate 852. Then use a last-digit check.
- Test candidate 915 using remainder nine, last digit and the bound that the sum is less than 900.
Questions 9–16: Other operations
- Calculate 247 × 36 by decomposition, then check its remainder-nine value.
- Does proposed product 8893 pass the check for Question 9?
- Does proposed product 8982 pass remainder-nine and last-digit checks? Is it actually correct?
- Test 327 as the difference in 803 − 476.
- Test proposed difference 337.
- Test proposed difference 372, then check exactly.
- Verify quotient thirteen and remainder one for 157 divided by 12.
- Test quotient sixteen and remainder one for the same division. Explain what the remainder-nine check misses.
Questions 17–24: Limits, decimals and toy codes
- For 157 divided by 12, a learner proposes quotient twelve and remainder thirteen. The reconstruction is exact. Why is the division answer still invalid?
- Can a remainder-nine check distinguish 348 from 834?
- Can it detect the fixed-position change from 102 to 192?
- Can it detect the fixed-position change from 462 to 472?
- Calculate 4.68 + 7.59 and write the corresponding whole-number check in hundredths.
- Calculate 1.2 × 0.3. Explain why a digit-sum match would not by itself reject 3.6.
- Under the toy rule in Section 9, append a check digit from zero through eight to payload 4826.
- Does 84267 pass that toy rule? Does a pass prove that the original payload was copied correctly?
12. Worked solutions
Solutions 1–8
1. Three. Add 4 + 6 + 2 = 12. Remove one group of nine to leave three. Place-value expansion shows that the original number and its digit sum have the same remainder.
2. Digital root nine; checking value zero. The digit sum is twenty-seven, which reduces to nine. But 999 is divisible by nine, so its remainder is zero. Keep the two conventions distinct.
3. Seven. The digits sum to 1 + 0 + 0 + 2 + 4 = 7. This is already in the allowed remainder range.
4. Five. The digit sum is 5 + 7 + 2 = 14. Fourteen is one group of nine plus five.
5. Passes. The inputs reduce to eight and seven, whose sum reduces to six. Candidate 825 also reduces to six. This establishes consistency with the check, not correctness by that check alone.
6. Rejected. Candidate 835 reduces to seven, but the inputs require checking value six. The unequal labels prove the stated equality is false, assuming the check was performed correctly.
7. Passes remainder nine but fails the last-digit check. Candidate 852 reduces to six. However, eight plus seven ends in five, so a total ending in two cannot be correct.
8. Passes the first two checks but fails the bound. Candidate 915 reduces to six and ends in five. However, the sum is less than 400 + 500 = 900, so 915 is impossible.
Solutions 9–16
9. 8892. Compute 247 × 30 = 7410 and 247 × 6 = 1482; add to obtain 8892. Its digit sum is twenty-seven, checking value zero, matching four times zero from the factors.
10. Rejected. The proposed 8893 has digit sum twenty-eight and checking value one. The factors require zero, so the candidate cannot be the correct product.
11. Both checks pass, but the answer is wrong. Candidate 8982 has checking value zero and ends in two. The exact decomposition gives 8892 instead. Their difference, ninety, is invisible to both remainder-nine and last-digit filters.
12. Passes and is exactly correct. Checking values give eight plus three reducing to two, matching 803. Separately, the exact inverse 476 + 327 = 803 verifies the difference.
13. Rejected. Candidate 337 reduces to four. Adding the subtrahend’s checking value eight gives twelve, reducing to three rather than the required two.
14. Passes remainder nine but fails exact reconstruction. Candidate 372 reduces to three, like the correct 327. Yet 476 + 372 = 848, not 803. The permutation of digits escaped the compressed check.
15. Valid. The exact identity is 157 = 12 × 13 + 1, and one is a valid remainder below twelve. The remainder-nine values also agree, but the exact identity plus the bound complete the verification.
16. Passes remainder nine but is false. Reconstructing gives 12 × 16 + 1 = 193. The difference from 157 is thirty-six, a multiple of nine. Therefore the filter misses this incorrect quotient.
Solutions 17–24
17. The remainder is too large. Although 12 × 12 + 13 = 157, a normal remainder after division by twelve must be less than twelve. Regroup another twelve to obtain quotient thirteen and remainder one.
18. No. Both numbers contain the same digits and have digit sum fifteen, reducing to six. A pure digit permutation cannot be detected by this check alone.
19. No. Replacing zero by nine in the tens place adds ninety. Both 102 and 192 reduce to three. The special zero-to-nine substitution is invisible modulo nine.
20. Yes. The checking value changes from three for 462 to four for 472. The difference ten is not divisible by nine, so the labels disagree.
21. 12.27. Use hundredths: 468 + 759 = 1227. Their checking values are zero plus three equalling three. Return the result to hundredths rather than treating 1227 as whole units.
22. 0.36. Twelve tenths times three tenths gives thirty-six hundredths. Moving the decimal point to obtain 3.6 does not change the raw digit sum, so a place-value or magnitude check is needed.
23. Seven. Payload digits sum to twenty. Adding seven reaches twenty-seven, a multiple of nine. The completed toy code is 48267.
24. It passes, but copying is not proved. The digits of 84267 also sum to twenty-seven. The swapped payload survives because the rule ignores order. A pass establishes only compliance with the toy condition.
13. Transfer: construct and defeat a check
For 586 + 279, find the exact sum and its remainder-nine checking value. Then test the deliberately incorrect candidates 874 and 955. Finally create a toy check digit for payload 5718.
Exact solution: 586 + 279 = 865. The input checking values are one and zero, so the sum should reduce to one. The digits of 865 sum to nineteen and reduce to one.
Candidate 874: its digits also sum to nineteen, so it passes remainder nine. The units calculation six plus nine ends in five, not four, rejecting it. Candidate 955: it reduces to one and ends in five, passing both filters. But 586 < 600 and 279 < 300 imply the sum is less than 900, rejecting 955.
Toy code: payload 5718 has digit sum twenty-one and checking value three. Append six to reach twenty-seven, giving 57186. A rearrangement of its payload digits would still pass, so the check cannot certify their order.
14. Teach the limitation as part of the method
Begin with a correct calculation and one candidate that fails. Then deliberately supply a wrong candidate that passes. Ask the learner to explain why the two results are compatible. This prevents the checking routine from being remembered as an unconditional proof method.
If a child says “it passes, therefore it is right”, return to numbers with the same remainder. Three and twelve have the same remainder on division by nine, but they are not equal. The arithmetic is simple enough to make the logical gap visible.
If the child is confused by digital root nine versus remainder zero, use only remainders until that convention is secure. Do not require two notational systems simultaneously. Once the learner understands zero as the remainder of a multiple of nine, the conventional digital-root label can be compared explicitly.
Match the check to the suspected mistake
A misplaced decimal point calls for magnitude and units. A digit substitution may be caught by remainder nine. A transposition calls for another positional or exact calculation. A valid arithmetic result attached to the wrong story quantity calls for rereading the model. The same check should not be prescribed for every error.
For broad checking habits, use Estimation, Reasoning and Checking. For the number structure behind remainders, use Factors, Multiples and Divisibility. This article’s distinct purpose is understanding a compressed check and its information loss.
A delayed return
After a gap, provide one candidate that fails remainder nine, one wrong candidate that passes, and one decimal-placement error. Ask the learner to choose an appropriate next check and explain what the result establishes. Correct vocabulary matters: rejected, not rejected, and exactly verified are three different conclusions.
No score or elapsed time is needed to make that observation useful. The strongest evidence of understanding is a learner who can create a passing wrong answer deliberately and explain why a fuller verification rejects it.
Sources and boundaries
University of Cambridge NRICH — Divisibility Tests, by Tim Rowland gives a related place-value explanation of divisibility tests, including the relationship between powers of ten and multiples of nine. The worked calculations, false candidates and toy code in this guide are independently developed teaching examples.
For official subject documents, consult the MOE Primary curriculum and syllabus page alongside the learner’s school programme. This is enrichment rather than a complete official syllabus or a prescribed examination checking technique.
Continue the worked-learning collection
Balance Scales and Weighing Puzzles examines what a comparison can distinguish. Intervals, Fence Posts and Cuts checks whether the calculation counts the right kind of object. Counting Rectangles, Squares and Triangles shows how a complete enumeration can justify a total.
Return to the BTT Primary Mathematics Learning Hub. A useful check should make its promise explicit: it may reject a candidate, narrow the possibilities, or verify a complete relationship. Knowing which promise it fulfils is part of the mathematics.