For this experiment I took 20 identical coins. I scooped them up into both hands and jangled them around to make them random before dumping them on the table.
I removed any that were "heads". They had "decayed and were no longer radioactive. I counted what was left.
Then I scooped up what was left and jangled them in both hands before dumping them on the table. I removed the heads and counted the rest and so on. For each throw, I wrote down how many I counted. In theory, you'd expect the number left to halve each throw. So we can't say which coins will be removed - it's random - but we can predict that half of them will be removed. You can't predict for an individual but you can predict for a large group. That's what half-life is for radioactivity.
In class I normally use 6-sided wooden cubes that I call dice. One side of each cube is marked. I do the same experiment but remove the ones that come out marked side up each time.
This time the chance of one being removed is much lower. It takes longer for the numbers to go down - more throws before the numbers halves. We'd say that the dice have a longer half-life than the coins.