I don't think that I ever even heard of completing the square until graduate school when I was forced to teach it along with a number of other semi-archaic things like synthetic division.
I think that you can make a case for completing the square as a technique for deriving the quadratic formula and possibly for answering the question "why do all quadratics have parabola graphs". Now, once you actually have those, there may not be much of a case for having them do it by scratch again and again. The quadratic formula really is quite effective; why not use the tools we are given?
The question of "why parabolas" is really given by one of the formulas you get along the way:
$$ ax^2+bx+c = a\left(x+\frac{b}{2a}\right)^2 - \frac{b^2-4ac}{4a}$$
and observing how it tells you that the graph must be a shift and scale of the standard $x^2$ graph. That's worthwhile and you can push the idea along to cubics! There, you have the notion of writing a general cubic in the form of a depressed cubic by a similar stretch and shift. The new, depressed cubic now looks like
$$ t^3 +pt+q. $$
Notice the lack of a quadratic term. You can now figure out what any kind of cubic graph looks like by playing around with $p$ and $q$.