A table has a shape before it has an answer
Two school clubs are buying materials for two different displays. A quantity table gives how many cards, labels and folders each club needs. A price table gives the cost of one item in each category. It is tempting to multiply every number that looks close to another number. That would be a mistake. The price of one card must be paired with the number of cards, not the number of folders. What gives the calculation its meaning is the alignment of the categories. Matrix multiplication is a compact way to preserve that alignment while performing several related calculations.
The words row and column are not decorative vocabulary. They tell the reader what one position means. Suppose a quantity matrix A has two rows, one per club, and three columns, one per kind of material. A matrix B contains the three unit prices in a single column. The product AB then contains two total costs in one column. The arithmetic works because the three material categories in A match the three prices in B. We should be able to describe the output before using a calculator. If we cannot, we have not yet decided what problem we are solving.
This guide takes the next step from Volume 0166, which owns basic matrix layout, reading and scalar multiplication. It also follows Volume 0301 on the difference between persuasive examples and supported conclusions. In both subjects, good examination work begins by identifying exactly what the evidence or number can support. This volume concentrates on the product of two matrices and the decisions that follow it, not on repeating the foundations.
Current official examination scope
The SEAB G3 SEC 2027 school-candidate syllabus directory, updated 26 October 2026, identifies Mathematics as K310, with 4052 as the earlier O-Level reference. The 2027 K310 Mathematics syllabus includes representing and interpreting matrices, multiplication by a scalar, and sums or products of two matrices where appropriate. Paper 1 and Paper 2 each last 2 hours 15 minutes, carry 90 marks and contribute 50%. The final Paper 2 question involves a real-world application. A learner does not need to imagine a separate matrix paper or predict which year a particular product will appear. The important skill is selecting and carrying out the appropriate operation, then interpreting its outcome.
Both papers permit approved calculators and require essential working. Unless a question directs otherwise, a non-exact numerical answer is normally given to three significant figures, and angles in degrees to one decimal place. Exact integer matrix products should remain exact. These exercises are original teaching tasks, not official specimen items. Their purpose is to build the mathematical thinking needed for a variety of possible SEC questions, including unfamiliar contexts.
The inside dimensions are permission; the outside dimensions are the destination
A matrix with two rows and three columns is called a 2 by 3 matrix. Let A be 2 by 3 and let B be 3 by 4. We can form AB because the three columns in A match the three rows in B. The product is 2 by 4. The shared dimension identifies the categories being combined; the remaining dimensions identify the output rows and columns. A 2 by 3 matrix multiplied by a 2 by 4 matrix is not permitted in that order. Neither a calculator nor a clever arrangement of brackets changes this rule.
This is different from adding matrices. Addition requires matrices of exactly the same size because the first entry is added to the first entry, the second to the second, and so on. Matrix multiplication joins a row of the first matrix with a column of the second. Those are different operations with different jobs. The student who memorises only that ‘the numbers have to match’ may apply the wrong rule. Instead, ask what is being paired: corresponding entries for addition, or a complete row and a complete column for a product.
How one entry of the product is formed
Take A = [[2,3,1],[4,0,2]], where the two rows represent two exhibitions and the three columns represent cards, labels and covers. Let B = [[5],[7],[4]] be their respective costs in dollars. Row one of A pairs with the column of B: 2 times 5, plus 3 times 7, plus 1 times 4, giving 35. Row two gives 4 times 5, plus 0 times 7, plus 2 times 4, giving 28. Thus AB = [[35],[28]]. The dimensions 2 by 1 are not an accident: there is one output cost for each of two exhibitions.
The entry 0 does not make a category disappear. It says the second exhibition uses no labels. We still keep the label position so that the cost of a cover is not mistakenly matched with the price of a label. Matrix position can matter even where a number contributes nothing. This is one reason a matrix is more useful than an unlabelled string of numbers: it records what each figure represents.
Now imagine two suppliers. Put the prices of supplier A in the first column of B and supplier B in the second. If B = [[5,6],[7,8],[4,3]], AB becomes [[35,39],[28,30]]. The first row compares the first exhibition’s total cost at both suppliers; the second compares the second exhibition. If a task requires each exhibition to buy from the cheapest supplier, the first exhibition should choose A at 35 and the second A at 28. If an additional restriction requires one supplier for the whole purchase, sum down each supplier’s column: A costs 63 and B costs 69. An apparently simple change in the question changes which total is appropriate.
Why a product is not generally reversible
For ordinary real numbers, 3 times 5 equals 5 times 3. For matrices, AB is not generally equal to BA. Sometimes one product is defined and the reverse is not. For example, a 2 by 3 matrix times a 3 by 4 matrix produces a 2 by 4 result. Reversing the order tries to multiply a 3 by 4 matrix by a 2 by 3 matrix. Four does not match two, so BA is undefined. Even when both products exist, their entries often differ. A learner must not transfer the commutative rule for numbers to matrices without testing it.
Consider C = [[1,2],[0,1]] and D = [[1,0],[3,1]]. CD = [[7,2],[3,1]], whereas DC = [[1,2],[3,7]]. Both products are two by two, yet they are not equal. The reason is structural. In the first product each row of C is paired with each column of D. In the second, each row of D is paired with each column of C. The pairs have changed. A memorised statement such as ‘multiply across’ is not enough. A student must specify whose row and whose column are being used.
Build an answer that can be checked
Use five small decisions. Name the meaning of each row and column. Write the dimensions beside each matrix. Check the matching inner dimensions. For each output entry, multiply the corresponding elements of one row and one column before adding. Then verify the shape, unit and interpretation of the completed product. Keep the individual working visible on paper. A calculator may confirm the arithmetic, but it cannot tell whether the chosen quantities were related properly.
A candidate who writes the correct four values in a transposed arrangement has lost information. So has a candidate who reports a cost in centimetres or a distance in dollars. The output must be checked as a meaningful result, not merely a numerical result. This is particularly important when the examination question asks for a choice, cost comparison, combined journey or table-based interpretation.
Worked examination clinic: one distinct decision per task
Every exercise below is original. Work through the task first, then read the reasoning and the final check. Some questions require computation; others ask whether a computation is legitimate. Those interpretation questions matter because the most damaging errors often occur before arithmetic begins.
01 Check the output dimensions
Task. A is 4 by 2 and B is 2 by 5. What is the shape of AB?
Worked reasoning. The inner dimensions are both two, so AB exists. Its result has four rows and five columns. There will be twenty entries, each made from two paired products.
Check. Do not conclude the result is two by two because two is the matching number. The common dimension tells us multiplication is permitted; the outer numbers describe the answer.
02 Recognise an illegal product
Task. A is 3 by 2 and B is 4 by 3. Decide whether AB or BA exists.
Worked reasoning. AB does not exist because its inner dimensions would be two and four. BA does exist: a 4 by 3 matrix times a 3 by 2 matrix yields a 4 by 2 result.
Check. Never label an impossible product as zero. A zero matrix is a valid output containing zero entries; undefined multiplication has no output.
03 One row and one column
Task. Compute [[2,4,1]] multiplied by [[3],[5],[7]].
Worked reasoning. The three positions in the row align with the three in the column. Compute 2×3 + 4×5 + 1×7 = 33. The answer is a one by one matrix whose entry is 33.
Check. Adding 2, 4, 1, 3, 5 and 7 would produce a different calculation. The pairing operation is what defines matrix multiplication.
04 A zero must stay in its position
Task. Compute [[0,6,2]] times [[9],[2],[5]].
Worked reasoning. The expression gives 0×9 + 6×2 + 2×5 = 22. The zero contributes nothing to this total but preserves the first material category in its proper place.
Check. Do not slide later entries into the first position after encountering zero. That would pair the wrong counts with the wrong rates.
05 Protect a negative contribution
Task. Compute [[3,−2]] times [[4],[5]].
Worked reasoning. The answer is 3×4 + (−2)×5 = 2. Negative entries may describe corrections, changes in balance or a signed mathematical quantity; their interpretation comes from the question.
Check. Treat the negative sign as part of the entry. Brackets around the negative product help prevent the common error of replacing −10 with +10.
06 Generate two output columns
Task. Let A = [[1,2]] and B = [[3,4],[5,6]]. Calculate AB.
Worked reasoning. The first column gives 1×3 + 2×5 = 13. The second gives 1×4 + 2×6 = 16. The product [[13,16]] has one row and two columns.
Check. The second output uses the second column of B, not a nonexistent second row of A. Rows stay fixed while output columns change.
07 Generate two output rows
Task. Let A = [[2,1],[3,4]] and B = [[5],[6]]. Calculate AB.
Worked reasoning. The first row gives 2×5 + 1×6 = 16. The second gives 3×5 + 4×6 = 39. The product [[16],[39]] carries one value for each original row.
Check. If the question asks for a matrix, stop at the two-row result. Adding the two outputs into one number would discard the distinction between cases.
08 Complete a square product
Task. Let A = [[1,2],[3,4]] and B = [[2,0],[1,5]]. Find AB.
Worked reasoning. Entry one is 1×2 + 2×1 = 4. Entry two is 1×0 + 2×5 = 10. The lower row is 3×2 + 4×1 = 10 and 3×0 + 4×5 = 20. AB is [[4,10],[10,20]].
Check. The two equal middle entries happen to match in this example. Do not assume a matrix product must have mirror-image entries across its diagonal.
09 Use the identity appropriately
Task. Find [[2,5],[7,1]] times [[1,0],[0,1]].
Worked reasoning. The second factor is the two by two identity. The product is [[2,5],[7,1]], unchanged. The identity acts as a neutral multiplier for a compatible square matrix.
Check. The identity property is a special pattern. It does not make multiplication by every diagonal matrix harmless or reversible.
10 Zero matrix is not a shortcut
Task. Compare multiplying a 2 by 3 matrix by a 3 by 2 zero matrix and by a 2 by 3 zero matrix.
Worked reasoning. The first product is legal and produces a 2 by 2 zero matrix. The second product is not legal because the inner dimensions are three and two.
Check. The values in a matrix cannot overrule its dimensions. A matrix full of zeroes has the same compatibility requirements as one full of positive numbers.
11 Compare suppliers without mixing columns
Task. Two clubs need [[4,2],[1,5]] units of paper and paint. Two supplier-price columns are [[3,5],[7,4]].
Worked reasoning. The cost matrix is [[26,28],[38,25]]. Club one costs 4×3 + 2×7 = 26 from the first shop, or 4×5 + 2×4 = 28 from the second. Club two costs 38 or 25 respectively.
Check. The cheapest shop depends on which club is buying. The first shop wins for club one, the second for club two; avoid a universal recommendation.
12 Choose addition instead of multiplication
Task. Two schools record the same three categories of donated books in matching 2 by 3 matrices. What operation combines their counts?
Worked reasoning. Add the matrices entry by entry, provided the row and column labels mean the same thing for each school. No conversion of categories is needed.
Check. The fact that some matrix product might be arithmetically possible does not make it the meaningful operation. Interpretation comes before computation.
13 Distinguish scalar from product
Task. A = [[2,3],[4,5]]. Explain why multiplying by scalar three is different from multiplying by an arbitrary 2 by 2 matrix.
Worked reasoning. Scalar multiplication makes every entry three times as large, giving [[6,9],[12,15]]. A general matrix product connects rows to columns and combines entries. It could change them in a different way.
Check. The special matrix 3I reproduces scalar multiplication, but this exception does not erase the distinction between a scalar and a general matrix.
14 Price list orientation
Task. Three classes each report boys and girls. The class matrix is 3 by 2; a weighting for boys and girls should produce one output per class.
Worked reasoning. Write weights in a 2 by 1 column. The resulting 3 by 1 product then has one weighted total for each class. A 1 by 2 row of weights would be incompatible.
Check. A price or weight list presented horizontally can be rewritten vertically when the question defines it that way. Do not silently transpose without saying what changed.
15 Write the product in its actual shape
Task. A is 2 by 3 and B is 3 by 1. A student obtains the correct two values but writes them side by side.
Worked reasoning. The product must be a 2 by 1 column. Present the values in two rows. The placement carries the meaning that each result belongs to a different source case.
Check. A correct unordered list is not a correct matrix. Make your row breaks and brackets legible.
16 Let units expose a bad model
Task. A table gives electricity used in kilowatt-hours and another gives the tariff in dollars per kilowatt-hour.
Worked reasoning. Multiplying compatible quantities produces cost in dollars, because kilowatt-hours multiplied by dollars per kilowatt-hour leaves dollars. This can be a useful real-world matrix product.
Check. If the first table already gives dollar costs, multiplying by the tariff again would be an invalid model even if the dimensions matched.
17 Show that order changes numbers
Task. A = [[1,1],[0,1]], B = [[1,0],[2,1]]. Find AB and BA.
Worked reasoning. AB = [[3,1],[2,1]] while BA = [[1,1],[2,3]]. Both products are defined and have the same shape, but the values differ. One example is sufficient to disprove a claim of universal equality.
Check. Checking dimensions is necessary but insufficient for assuming two defined products are equal.
18 Turn a matrix entry into an equation
Task. A = [[x,2]], B = [[3],[4]], and AB = [[17]]. Find x.
Worked reasoning. The row-column multiplication gives 3x + 8 = 17, hence x = 3. Substitute it to confirm that 9 + 8 equals 17. A matrix equation may reduce to familiar algebra.
Check. There is no need to use an advanced inverse-matrix method when one entry creates a simple equation.
19 Use the known column as a check
Task. A = [[1,3]], B = [[2,5],[4,y]], AB = [[14,23]]. Find y.
Worked reasoning. The first output is 1×2 + 3×4 = 14, so our orientation is consistent. The second gives 5 + 3y = 23. Therefore y = 6.
Check. Confirm known entries before solving unknown ones. An incorrect orientation can otherwise produce a plausible but irrelevant equation.
20 Three input categories
Task. A = [[1,0,2],[3,1,1]] and B = [[2,1],[4,0],[5,3]]. Find AB.
Worked reasoning. The product is [[12,7],[15,6]]. For instance, row two with column one is 3×2 + 1×4 + 1×5 = 15. Four outputs each require three paired products.
Check. Do not conclude that rectangular factors must give a rectangular output. The outer dimensions here are both two.
21 Interpret a falsely described result
Task. A 2 by 4 matrix multiplies a 4 by 3 matrix. Someone says the answer shows four scores for each of three teams.
Worked reasoning. That interpretation describes a 3 by 4 table, not the actual 2 by 3 product. The calculation produces three output categories for each of two source cases.
Check. A shape check can disprove an interpretation without inspecting any entry. Use it before lengthy explanations.
22 Apply a common discount at the right stage
Task. A quantity matrix times a price matrix produces costs. Every final bill is reduced by ten percent.
Worked reasoning. Multiply the cost matrix by 0.9. This is justified because the same discount applies to every output entry. If different items received different discounts, the adjustment would belong earlier in the model.
Check. Do not transfer a single rule to situations with category-specific conditions. Where the percentage is applied matters.
23 Recognise missing information
Task. Two display costs are known after multiplying a quantity matrix by a price column. Does that always reveal the three original prices?
Worked reasoning. No. Two totals may give only two independent equations for three unknown prices. Without sufficient other information, multiple price lists could be consistent with those totals.
Check. A product can be uniquely calculated even when its original factors cannot be reconstructed. Distinguish forward computation from inverse inference.
24 Return to the purchasing restriction
Task. The product cost matrix is [[26,28],[38,25]]. Both clubs must use the same shop. Which is cheaper?
Worked reasoning. Add each shop column: first shop totals 64, second 53. The second shop is cheaper under the shared-supplier condition. If separate suppliers were allowed, the cheapest pair would total 26 + 25 = 51.
Check. The same correct matrix permits different decisions under different constraints. Read the decision question after the arithmetic.
25 Do not turn a special case into a law
Task. A student finds a particular pair of matrices where AB = BA and announces that every two matrices commute.
Worked reasoning. The conclusion is invalid. Some pairs commute, for example a compatible identity with another square matrix, but other pairs do not. Our earlier counterexample is enough to reject the universal claim.
Check. The words always and never should trigger a proof or counterexample check, not an intuition based on one example.
26 Label the common category
Task. A has days as rows and pieces of equipment as columns. B has equipment as rows and rental fee categories as columns.
Worked reasoning. The shared equipment categories allow the row-column pairing. The product gives fee-related outputs by day. Keeping the equipment order identical across the two factors is necessary.
Check. A matrix may be numerically compatible but semantically wrong if one table lists helmets before bicycles while the other reverses that order.
27 Check a tight threshold
Task. A matrix product gives total equipment costs of $117.50. A school has a $120 limit. Write the decision.
Worked reasoning. The proposed purchase fits the stated budget by $2.50. An answer that says only ‘yes’ hides the calculation and the small remaining margin. State the value, unit and decision.
Check. Do not round $117.50 to $120 before checking whether it stays below the limit. The exact comparison matters.
28 A total across several categories
Task. A 2 by 3 quantity matrix times a 3 by 2 price matrix gives a 2 by 2 table. The task then asks for the grand total across both suppliers.
Worked reasoning. It is necessary to clarify whether the two supplier columns are alternatives or costs incurred together. If they are competing offers, adding both columns double-counts an alternative. If they represent two distinct expense categories, addition may be appropriate.
Check. A mathematical operation can be correct while the interpretation is wrong. Ask what each final column represents before summing.
29 Recognise a matrix expression with three factors
Task. A 2 by 3 quantity matrix A, a 3 by 2 conversion matrix B and a 2 by 1 final weighting matrix C are given.
Worked reasoning. AB gives a 2 by 2 matrix, and then (AB)C gives 2 by 1. The dimensions also permit B C first: B is 3 by 2 and C is 2 by 1, giving 3 by 1, then A(BC) gives 2 by 1. The associative rule preserves the result when products are defined.
Check. Associativity does not mean commutativity. Rearranging brackets without reversing factor order is different from swapping B and C.
30 Keep a symbolic output intelligible
Task. A = [[a,b]], B = [[x],[y]]. State the product and interpret it if a and b are quantities and x and y are prices.
Worked reasoning. The product is [[ax + by]]. Both terms have the same unit, money, so their sum is meaningful. If one output were time and the other money, addition would not be physically interpretable without a stated conversion.
Check. A variable is not less important than a number. Its unit and meaning remain part of the expression.
31 Audit a matrix claim in prose
Task. A student writes: ‘Because there are three suppliers, the result must have three rows.’ The supplier matrix has three columns.
Worked reasoning. The supplier label actually determines the output columns if suppliers are represented by columns of the right-hand matrix. Row and column identity come from the original arrangement, not from the names alone.
Check. Never infer layout from a noun without checking which axis holds it. A good explanation names both axis and category.
32 Prove why a missing value cannot be negative
Task. A = [[2,x]] is a quantity matrix of nonnegative item counts, and B = [[3],[4]]; the product is [[2]].
Worked reasoning. The equation is 6 + 4x = 2, so algebra gives x = −1. That solution contradicts the premise that x is a nonnegative count. The correct model conclusion is that the supplied data are inconsistent.
Check. An algebraically valid answer can be inadmissible in context. Record the contradiction instead of reporting minus one item.
Move from guided questions to independent performance
In the first practice session, write the sizes of the input matrices beside every question. Stop before any arithmetic and say whether the product is legal and what its output size must be. Once you achieve this accurately, introduce the row-column calculations. This removes two sources of confusion at once. A student learning the shape rule should not also be racing through negative-number arithmetic and unfamiliar real-world units.
In a second session, mix matrix addition, scalar multiplication and products rather than grouping ten identical questions together. When every exercise in a worksheet is of the same operation, students may learn to repeat the operation without noticing the information that calls for it. A mixed set makes selection part of the task. For every answer, write a short reason, such as ‘the columns of the first matrix match the rows of the second’ or ‘the two inventories have matching labels, so entrywise addition applies’.
In the third session, remove the labels temporarily and ask the learner to reconstruct them. Imagine a school timetable, a material price list and several class requirements. How should the data be arranged to produce one cost per class? How would it change if the output must compare two possible suppliers? This transfer task is more demanding than multiplying the numbers of a prepared example, but it resembles the challenge of turning prose into Mathematics.
Repair errors according to their cause
Keep a small log of the first incorrect step, not just the final wrong answer. If the inner dimensions did not match, practice dimension compatibility. If the correct row was paired with the wrong column, practise tracing one output entry with a finger across and down. If a zero made an entry shift position, reinforce the labels. If the calculated product was correct but the recommendation was wrong, change the practice to real-world modelling and decision constraints. One vague instruction to ‘be more careful’ is much less useful than a repair matched to a particular failure.
The Secondary Mathematics deliberate-practice guide explains this focus on one bottleneck. Students moving from primary school can also revisit PSLE Learner’s Guide Vol 0078 on totals and group sizes. That earlier habit—keep the right quantity attached to the right group—is exactly what matrix labels protect at secondary level.
An examination-room finish
When time is limited, the safest shortcut is clarity. Write the matrix shapes. Circle the matching dimensions mentally. Form one product entry from a complete row and a complete column. Keep the result arranged in its correct dimensions. Finally, read the question’s decision constraint and write an interpretation in the correct units. The purpose is not to produce a decorative rectangle of numbers. It is to show how labelled information leads to a defensible result.
The next volume uses the same discipline in Chemistry. Instead of rows and columns matching categories, a laboratory question asks which observable change belongs to which test, what it identifies, and what it cannot prove. In both subjects, accurate performance comes from protecting relationships rather than merely recognising familiar words.
Official sources and connected learning
SEAB 2027 G3 syllabus directory · K310 Mathematics syllabus · Volume 0168 on matrix-based charging decisions · After-PSLE learning route.
