A shop records hoodies and T-shirts sold in sizes S and L. Rows are hoodies and T-shirts, columns are S and L: June is J=[1220815] and July is K=[1025911]. The total for both months is J+K=[22451726]; each entry adds the two numbers in the same position. The change from June to July is K−J=[−251−4], and the −4 shows that four fewer large T-shirts sold.
Adding position by position only makes sense because the two tables share a layout. If July's table had a third column for size M, June's table would have nothing to add to it, so matrices of different sizes cannot be added or subtracted.
A scalar is a single number that multiplies every entry. If a stock table Q counts boxes and each box holds 6 items, then 6Q counts items, and every entry is six times as large. Familiar number rules carry over: A+B=B+A, 2(A+B)=2A+2B, and A−A is the zero matrix.
Multiplying two matrices is a different operation. AB is not found by multiplying matching entries; each entry of AB combines a whole row of A with a whole column of B. So 2A and AB follow different rules, even though both are called multiplication.
Think of it like this. Adding matrices is like laying two transparent spreadsheets with identical grids on top of each other and adding the two numbers in each cell. Scaling is one dial that multiplies every cell at once. If the grids are different sizes the sheets cannot line up, and the sum does not exist.
Core idea. Add or subtract matrices entry by entry (same dimensions required), and scalar multiplication scales every entry.