Give a robot standing at (2,3) two instructions: move 4 right, then make a quarter turn counterclockwise about the origin. After the first, it is at (6,3). The second acts on where the robot is now, not where it started: a 90° counterclockwise rotation sends (x,y) to (−y,x), so (6,3) goes to (−3,6).
Now swap the instructions. Rotating (2,3) first gives (−3,2), and moving 4 right gives (1,2). Same two moves, a different final spot, because the rotation turns about a fixed point, the origin, and the translation changes how far the robot is from it. That is what not commutative means: the order changes the result, the way putting on socks then shoes differs from shoes then socks.
Some pairs do commute, and they can fool you. Translating right 4 and reflecting over the x-axis give the same result in either order, because a horizontal slide never changes a point's distance from the x-axis. Translating up 4 does not commute with that reflection: (2,3) goes up to (2,7) and reflects to (2,−7), but reflected first it goes to (2,−3) and then up to (2,1). One matching example proves nothing about order in general.
In the notation T2∘T1(P)=T2(T1(P)), the transformation written nearest the point acts first, just as in f(g(x)) you work out g first.
Think of it like this. Chaining transformations works like following a recipe: whisk the eggs, then fry them, and each step acts on what the last step produced. Swap the order, fry then whisk, and you get something else. Some steps, like adding salt and adding pepper, can go in either order, just as any two translations always commute.
Core idea. Composing transformations applies them in sequence, and the order usually changes the outcome.