Oddly, I don't find WA to be great. WA can do complex math, but it's slow. To give a good example of what I've come to expect from all three, type "2 feet + 2 meters" into:

* WA: You'll get a correct result, but it takes 4 seconds.

* Google: 0.185806 square meters, obviously (instantly)

* DDG: Search results

Curiously, I find Google's to be the most helpful. I see the obviously incorrect result and instantly fix it. DDG gives me nothing. WA is too slow to e.g. play around with numbers.

My last computation was trying to figure out which bolt to get for a hole (with McMaster's selections for both bolts and, at some level, holes). I was converting hole sizes / bolt sizes between metric and imperial, and between fraction and decimals, playing around until I found a combo which worked. Once I had Google giving the right answers, I could play around until it worked. With WA, I would have had to wait 4 seconds each time.

By the time I'm doing anything beyond that, it's generally in numpy. There isn't really a spot for WA in between.