I realize proofs for this identity have been asked before, but the ones I've found so far involve either induction, generating functions, counting number of functions, etc. I've yet to find this precise argument which I'm struggling a bit with. At any rate, apologies if this counts as a duplicate post.
Again, the claims is,
$$n! = n^r - n(n - 1)^r + \frac{n(n-1)}{2!}\left(n -2 \right)^r - \dots \ (\text{when} \ r = n) $$
Here is the proof:
We have,
$$ (e^x - 1)^n = \left( x + \frac{x^2}{2!} + \frac{x^3}{3!} + \dots \right)^n \tag{1} $$
Also, by the binomial theorem we have,
$$ (e^x - 1)^n =e^{nx} - ne^{(n-1)x} + \frac{n(n-1)}{2!}e^{(n-2)x} - \dots \tag{2}$$
Expanding each of the terms $e^{nx}, e^{(n-1)x}, \dots $ we see that the coefficient of $x^r$ in (2) is
$$ \frac{n^r}{r!} - n\frac{(n-1)^r}{r!} + \frac{n(n-1)}{2!}\left( \frac{(n - 2)^r}{r!} \right) - \frac{n(n-1)(n-2)}{3!} \left( \frac{(n-3)^r}{r!} \right) + \dots \tag{3}$$
Now equate coefficients of $x^r$ in (1) and (2) and the result follows.
\begin{align} \\ \end{align}
The logic between $(2)$ and $(3)$ isn't quite clicking. "[E]xpanding each of the terms" to me would mean something like $e^{rx} = 1 + \frac{(rx)^2}{2!} + \dots $ for the $r^{\text{th}}$ term. It's unclear to me if we're supposed to expand every binomial term in $(2)$ into a series and then add them all up or how that would produce an $x^r$ term if we did.
If someone could slap some training wheels on the derivation of $(3)$ I'd be much obliged.