Most of the “rules” of writing are merely guidelines to point you in the right (write?) direction. Truthfully, there’s only one that you absolutely must follow:
You can do anything you want — as long as it works.
The first part is simple; it’s just what you wanted to hear. The second part, there’s the rub. How, exactly, do you go about figuring out if something works?
My first suggestion is to follow the “rules.” They’re not there to constrain you. They exist as helpful guides to keep you from screwing up. If you’re going by the acceptable standards, you’re probably going to be okay.
Sometimes, you simply have to be true to your artistic vision, or, maybe, you just want to be contrary. Basically, sometimes you just gotta break all the rules. In that case:
Use your discernment as a writer. There is absolutely no one better qualified than you to determine if your vision is being translated onto the page because no one else knows your vision. With experience comes discernment. Flag any areas where you felt you’ve went off the beaten path. Continue writing. Once you’ve gained some distance and perspective, go back to those sections and ask yourself, “Did I really accomplish what I wanted?”
The problem is that, while you are the best person to determine if you’ve translated the story in your head correctly, you’re also the worst. You know exactly what you meant to say, and your mind will trick you into reading what you meant instead of what you wrote. In that case:
Find good beta readers. In the absence of good ones, bad ones will work. Remember, however, the cardinal rule of dealing with such, good or bad, as the old saying goes — they’re usually right when they say you’ve screwed up but usually wrong in telling you how to fix it. What I’ve found is that there will be long stretches of text with no comments. Then, they’ll be a section where multiple readers have placed a comment. These remarks may be sentences or paragraphs or even half a page apart. They may critique different things entirely. One may question my word choice while another mentions character. The thing I take away is that the scene didn’t work, and I need to fix it. And not necessarily by changing either of the things the commenters brought up.
The problem with beta readers is that it’s a bit like the blind leading the blind, and it’s sometimes hard to trust them completely. That’s why you need an expert.
Pay an editor who you trust. For one thing, the fact that you’re laying out cold, hard cash gives the comments you get back instantly more weight. What you get for free (or even as an exchange) is not nearly as valuable as what you pried open your wallet for. Your editor should be the most experienced expert you can find and afford. Don’t skimp on this step. He’ll be your best friend in that he’ll take your work to the next level. He’ll be your worst enemy in that he’ll see all the flaws you hoped you had hidden.