I don't think you can judge a head coach based on the outcome of any one game. The northwestern game is a perfect example of why: football is not a tiny little carve out that's immune to the rules of probability and statistics. Sometimes insanely improbable things (Cam McDaniel fumble) happen. Play calls are a matter of balancing risk and reward: sometimes risky plays pay off, sometimes they fail spectacularly. A coach's impact on one single game is very subject to the vagaries of statistics.
To a lesser degree, the same can be said of a single season. Look at any sport where teams play more than 12 games. They all go through streaks. That doesn't mean the team or the coaching gets worse or better, it just means that the outcome of athletic contests are very rarely 100% certain so there's bound to be some fluctuation over small sample sizes. Football, again, is not some odd carve out that doesn't adhere to the rules of probability and the college football season is a very small sample size.
That being said, you can certainly evaluate a coach over 5 years, and you can use a single game (or sing play) as examples of broader tendencies. I think Kelly does have some weaknesses that can't just be dismissed at this point: consistent problems with his secondary, consistent problems with red zone offense, generally weak special teams (though that may change this year). I don't think those problems outweigh the good he's done for the program. He needs to continue to improve as a coach, but I'm excited about the direction this program is heading, and that's good enough for me.