Akshay DineshMatrix multiplication can look like a strange rule at first: rows meet columns, numbers are paired and added, and the answer often seems less obvious than ordinary multiplication. The rule starts to make more sense when a matrix is treated as a move. One matrix can stretch a shape, another can rotate it, and their product can describe what happens when those moves happen one after the other.
That idea is one of the reasons matrix multiplication appears so often in linear algebra, computer graphics, data work, physics, engineering, and systems of equations. It is not just a way to fill in a table of numbers. It is a compact language for combining actions. Once that purpose is clear, the row-by-column process stops feeling arbitrary and starts looking like careful bookkeeping.
A matrix can act like a rule for moving vectors
In linear algebra, a vector can represent a point, a direction, a force, a color value, a data record, or any object described by ordered numbers. A matrix can act on that vector by turning the input numbers into new output numbers. For example, a two-by-two matrix can take a vector such as \((x, y)\) and produce a new vector \((x’, y’)\).
One matrix might double every horizontal distance. Another might swap horizontal and vertical components. Another might shear a grid so that vertical lines lean while horizontal lines stay level. These are different moves, but they share an important feature: straight grid lines stay straight, and the origin stays fixed. That is why they are called linear transformations.
The columns of a matrix give a useful visual clue. In a two-dimensional matrix, the first column shows where the basic horizontal vector lands, and the second column shows where the basic vertical vector lands. Once those two destinations are known, every other vector can be rebuilt from them. A vector such as \((3, 2)\) means three copies of the horizontal basis vector plus two copies of the vertical basis vector, so the matrix transforms it by moving those basis pieces and adding the results.
Multiplication keeps track of two moves in order
Suppose matrix \(B\) acts on a vector first, and matrix \(A\) acts after that. Written in symbols, the two-step process looks like \(A(B\mathbf{v})\). The product \(AB\) is the single matrix that does the same combined job. It sends every vector exactly where it would land after first applying \(B\), then applying \(A\).
This is why the order matters. The expression \(AB\) means that \(B\) touches the vector first, even though \(A\) appears on the left. That feels backwards until the expression is read from the vector outward: start with \(\mathbf{v}\), apply \(B\), then apply \(A\). In many cases, doing the same two moves in the opposite order gives a different result.
A simple physical example helps. Imagine rotating a shape and then stretching it horizontally. Now imagine stretching it horizontally first and rotating it afterward. The final shape may not land in the same place or point in the same direction. Matrix multiplication preserves that difference, which is why \(AB\) and \(BA\) are usually not interchangeable.
The row-by-column rule is not random
The standard multiplication rule says that each entry of the product comes from one row of the left matrix and one column of the right matrix. If
\(A = \begin{bmatrix} a & b \\ c & d \end{bmatrix}\) and \(B = \begin{bmatrix} e & f \\ g & h \end{bmatrix}\), then
\(AB = \begin{bmatrix} ae + bg & af + bh \\ ce + dg & cf + dh \end{bmatrix}\).
Those four entries are not chosen for decoration. Each column of \(B\) tells where one basic input direction goes after the first move. Then \(A\) transforms those new column vectors. The first column of \(AB\) is \(A\) applied to the first column of \(B\), and the second column of \(AB\) is \(A\) applied to the second column of \(B\). The row-by-column rule is the arithmetic that makes those transformed columns come out correctly.
That also explains why the inner dimensions must match. A matrix can only act on vectors with the right number of components. If a matrix has three columns, it expects a three-component input. When multiplying two matrices, the right matrix produces outputs that the left matrix must be able to accept. The size rule is really an input-output rule in disguise.
A small example shows the combined move
Take two matrices:
\(B = \begin{bmatrix} 2 & 0 \\ 0 & 1 \end{bmatrix}\) and \(A = \begin{bmatrix} 1 & 1 \\ 0 & 1 \end{bmatrix}\).
The matrix \(B\) doubles the horizontal component and leaves the vertical component alone. The matrix \(A\) shears a vector by adding the vertical component into the horizontal component. If a vector starts as \(\mathbf{v} = \begin{bmatrix} 3 \\ 4 \end{bmatrix}\), applying \(B\) first gives \(\begin{bmatrix} 6 \\ 4 \end{bmatrix}\). Applying \(A\) after that gives \(\begin{bmatrix} 10 \\ 4 \end{bmatrix}\).
Now multiply the matrices directly:
\(AB = \begin{bmatrix} 1 & 1 \\ 0 & 1 \end{bmatrix}\begin{bmatrix} 2 & 0 \\ 0 & 1 \end{bmatrix} = \begin{bmatrix} 2 & 1 \\ 0 & 1 \end{bmatrix}\).
Using this product on the original vector gives
\(\begin{bmatrix} 2 & 1 \\ 0 & 1 \end{bmatrix}\begin{bmatrix} 3 \\ 4 \end{bmatrix} = \begin{bmatrix} 10 \\ 4 \end{bmatrix}\).
The same answer appears because the product matrix has stored the two-step action as one step. That is the main power of matrix multiplication: it lets a chain of transformations become a single transformation that can be reused again and again.
Where this idea shows up beyond homework
Computer graphics use this idea constantly. Moving a character, scaling a model, rotating a camera view, and projecting a 3D scene onto a screen can all be represented with matrices. Software can multiply those matrices together so that many small actions become one efficient transformation before thousands or millions of points are drawn.
The same logic appears in robotics and engineering. A robot arm may have several joints, each with its own rotation or shift. The position of the hand depends on combining those joint transformations in the correct order. In data science and machine learning, matrix multiplication can combine layers of numerical operations, moving information from one representation to another.
Systems of equations also benefit from the same structure. A matrix can organize the coefficients of several equations, while a vector stores the unknowns. Multiplication then expresses all equations at once. Instead of writing each equation separately, the relationship can be written as \(A\mathbf{x} = \mathbf{b}\), which keeps the system organized and makes larger problems easier to handle.
Common mistakes come from forgetting the meaning
Many mistakes with matrix multiplication happen when the procedure is memorized without the input-output idea behind it. A student may try to multiply matrices in either order, even when the dimensions do not fit. Another may assume \(AB = BA\) because ordinary numbers commute. Matrices are different because they often represent actions, and actions can depend strongly on order.
It also helps to remember that each entry in the product is a dot product between a row and a column. The row asks how one output component is calculated. The column supplies one transformed input direction from the previous matrix. Their product-and-sum calculation tells how much that input direction contributes to that output component.
Matrix multiplication becomes much easier when the arithmetic is tied to the picture: one matrix moves vectors, a second matrix moves the results, and the product records the combined move. The row-by-column rule is still worth practicing, but it is not the whole story. The real idea is composition, which means building one action from actions that happen in sequence.
Seen that way, multiplying matrices is not just a technical skill. It is a way to describe change clearly. Whether the subject is a grid on paper, a 3D object on a screen, a robot arm, or a large system of equations, the product of two matrices answers the same practical question: what single rule describes what happens after these moves are performed in order?
