Multiply two ordinary numbers and the order never changes the answer. Multiply two matrices instead, and swapping the order can hand back two completely different results.
The rule everyone assumes still applies
For real numbers, fractions, even complex numbers, multiplication is commutative: 3 times 5 gives exactly what 5 times 3 gives. It feels like one of the most obvious laws of arithmetic, so obvious it barely gets a name in school.
A matrix looks harmless enough to inherit that same habit. It is just a rectangular grid of numbers arranged in rows and columns, sized by how many rows (m) and columns (n) it has. Add two matrices of the same size and the process behaves exactly as expected, adding the numbers sitting in matching positions, one by one.
Where the resemblance quietly breaks
Multiplication tells a different story. Matrix multiplication does not just combine numbers, it represents composing linear transformations, stacking one geometric operation on top of another. Composing operations in one order rarely produces the same outcome as composing them in reverse.
That single difference means multiplying matrix A by matrix B is, in general, not equal to multiplying B by A. Two perfectly ordinary matrices can swap places and hand back two entirely different grids of numbers.
A simple example shows the mechanism at work. Multiply the row of numbers 2, 3, 4 by the column 1000, 100, 10, and the result is 2340, found by adding 2 times 1000, 3 times 100 and 4 times 10. It is essentially the same trick used to split an ordinary number into its hundreds, tens and units.
A rule shaped centuries before it had a name
Grids of numbers organized in rows and columns to solve systems of equations already existed in ancient Chinese mathematics, long before Europe formalized anything close to modern matrix theory.
The pieces came together slowly. In the 17th century, Gottfried Wilhelm Leibniz worked out an explicit formula for what would later be called the determinant, a single number calculated from a square grid that reveals whether the grid can be reversed at all. In the following century, Gabriel Cramer built a rule that solves an entire system of equations directly from the determinants of a table of coefficients. Pierre-Simon Laplace then found a way to break that same calculation down recursively, splitting a large grid into smaller ones, a method still taught in classrooms today.
A grid that can bend an entire shape
Numbers arranged this way do more than sit still. A 2 by 2 matrix representing a linear transformation can take a plain square and stretch it into a parallelogram, moving each of its four corners according to the matrix’s own numbers.
Flip a matrix on its diagonal, turning its rows into columns, and you get its transpose: the grid with rows 1, 2, 3 and 0, -6, 7 becomes a new grid with rows 1, 0 then 2, -6 then 3, 7. It looks like a simple rearrangement, but it carries deep consequences throughout linear algebra.
The size myth that trips up beginners
A common assumption holds that two matrices can only be multiplied if they share the exact same dimensions. The real requirement is narrower: the number of columns in the first matrix only has to match the number of rows in the second. A grid with 2 rows and 3 columns multiplies perfectly well with a grid of 3 rows and 5 columns, even though neither their shapes nor their overall size match at all.
Matrix multiplication does keep some familiar properties. It stays associative and distributive, just like ordinary numbers. The one rule it drops, quietly and without exception, is the one that says order does not matter.
A table of numbers that outgrew algebra
What started as a rectangular table for solving systems of equations has since spread far beyond algebra class, into graph theory, combinatorics and statistics. A tool built to organize numbers in rows and columns turned out to describe motion, stretch entire shapes, and solve problems no single equation ever could, one broken rule at a time.

