Five friends stand on a number line at 2, 4, 6, 8 and 10. Their mean position is 6. Measure how far each friend is from 6: 4 steps, 2, 0, 2 and 4. The friend at 2 is on the left and the friend at 10 is on the right, but both are 4 steps away. The average of these distances is 512=2.4, so a typical friend stands about 2.4 steps from the centre.
The absolute value is what makes this work. If you kept the signs (−4,−2,0,2,4), the left and right sides would cancel and the total would be 0. That happens for every data set, because the mean is exactly the balance point of the data. Distances ignore direction, so they add up instead of cancelling.
Compare 4, 5, 6, 7, 8 with 2, 4, 6, 8, 10: both have mean 6, but their MADs are 1.2 and 2.4, so the second set is twice as spread out. Range also measures spread, but it uses only the two end values and ignores the middle: 2, 6, 6, 6, 10 and 2, 2, 6, 10, 10 both have range 8, yet their MADs are 1.6 and 3.2.
Think of it like this. MAD is like the average walk from home to school for a class. Whether a student lives north or south of the school does not matter, only how far. Average everyone's walk and you know how spread out the neighbourhood is. One difference: in MAD the 'school' always stands exactly at the mean of the data.
Core idea. MAD is the mean of the absolute distances of each value from the data's mean.