While it's uncontroversial that reading certain books may be of enormous benefit, it seems that the word 'certain' in the foregoing is too often understated, and that the effort and intelligence one puts into deciding which things to read are of similar importance to the time and effort put into the reading itself.
Further, given the astronomical number of books we now have to choose from, sorting through them all to discover the tiny subset you would most benefit by reading in a given year (let's say) is a very difficult problem to solve well.
One aspect of the problem is that its ideal solution depends on your goals, so there is no one-size-fits-all answer. That said, I think there is one goal very often shared across readers, and the heuristics I'll list here relate to it. The goal could be called, 'forming an accurate picture of the world'; the books which are supposed to serve this goal effectively are often called 'important books'. However, there are now so many books which will be labelled 'important' by one authority or another that it's no longer prudent to go ahead and decide to read a book just because it frequently receives this labelling.
The following list contains the best heuristics I know for selecting important books to read—though I'm still interested in improving it, so please share if you've got some ideas on the subject. And, before we start, there's an important related principle found in a certain interpretation of a quote from Francis Bacon that I'd like to bring up: "some books are to be tasted, others to be swallowed, and some few to be chewed and digested". In other words, once you've decided to read something, keep an open mind to dropping it. You try to collect evidence that it's worth reading beforehand, but actually reading some of it is a very good supply of new information. And, one book may be worth skimming and getting the gist of, while another should be read with one's full attention and willingness to contemplate.
Anyway, the list:
1) Insist on clarity. I believe this is one of the most overlooked criteria when evaluating books. It may be that encoded somewhere in a work's obscure, ambiguous, or overly ornate phrasing there are riches to be found—but in my experience that outcome is the exception rather than the rule. Authors with justified confidence in the value of their work tend not to shy away from clarity. Certainly there are cases where the author just lacks the ability to place themselves in their readers' shoes, or whose focus on the aesthetic value of their work leads them to make compromises in other areas—so you've got to ask yourself if it's really worth it, considering all the other things you could read, to put the additional time and effort into figuring out what it is they're saying.
I hope I won't be mistaken here for saying to avoid difficult works, rather than unclear works. A work may be difficult because it was written unclearly, but it also may be difficult because of legitimate factors probably relating to a disparity between the background knowledge assumed by the work and your actual background knowledge.
It's a skill worth acquiring to test statements for clarity: if you read a passage and don't understand it, pin down the specific sources of confusion. If the author has been clear then it will come down to unfamiliar terms (or concepts referred to by those terms), and after investigating the troublesome terms the statement should become clear. If you know all the terms and yet still don't understand, you'll probably spot areas where the author is overly ambiguous, inaccurate, unjustified, or leaving out steps.
I'd recommend reading some random samples from candidate books and testing for clarity before going ahead.
2) Investigate the impact of its publication. The ability of readers to determine a book's importance is imperfect; both false positives and false negatives abound. In order to filter out more false positives, look for evidence outside individual opinions that the ideas from a book have some use. But keep in mind: the utility of this is in proportion to the width of the book's readership. And with that in mind:
3) Look for potentially overlooked works. If the work is legitimately important and it's been out for quite a while, chances are you've already indirectly incorporated many of its lessons; therefore, some of the highest benefit-yielding books will be relatively obscure.
What would cause a legitimate work to end up obscure? Here are a few ideas: They were writing at a time when their ideas were strongly against the current of accepted thought—and that time may be the present; their personality was not suited to self-promotion; they wrote in such a way that their readers were not often able to understand them (not because they lacked clarity, but because the set of ideas they drew from is unfamiliar to others in their field, for instance; so this is probably especially possible with strongly interdisciplinary thinkers); the book's subject matter is more important to you personally than it likely would be to more typical readers.
4) Count it as 'points against' if the book is very new. If the subject is especially germane to your interests and you're short on material from the field, then it probably makes sense to go ahead—but otherwise, let other readers do the filtering for you, and see if it's still 'important' in five or ten years (consider Wolfram's "A New Kind of Science," for instance). It's true you may miss something important this way, but unless you're extremely caught up on the older writings, you're also more likely to waste your time this way. The 'ooh, shiny new thing!' effect may also unduly bias your interest to recently released works, so take that into account.
5) Watch out for idolization of older works. Often times a classic's merit derives from its relation to the time in which it was written: at the time it was produced it may have been so earth-shattering that we still mostly hear echos of its original praise, rather than comments on its real worth to present day readers. (The situation is different however if your primary interest is in understanding the related historical era.)
If the book's ideas were of lasting consequence, it's a near certainty that more clear statements of them have been produced in the intervening years. You may lose some of the aesthetics of the original—but often times the true aesthetic value is debatable (or at least relative), and if it's truly there it still may not be worth losing the increased clarity and further refinement of ideas found in later works. (As an example of this which will work for a few readers: compare ancient and contemporaneous presentations of Buddhist ideas. For example, you might compare descriptions of core principles given in The Diamond Sutra, to those given by Culadasa or Ajahn Brahm; or even comparing D.T. Suzuki to Watts in 'The Way of Zen' is informative, though other factors of course play into the disparity there).
6) Investigate the author for likelihood of reliability, capability, integrity, and interestingness. I'd even go as far as reading a biography on them to get an idea of their character before committing myself to reading a demanding work of theirs (I'm tempted to do this with Foucault at the moment, for instance, because he sounds totally untrustworthy from what I've ascertained so far—yet he's extremely revered). There are of course limitations to this: genius assholes do exist; fearful and self-deprecating geniuses exist; incredibly interesting internal lives are sometimes paired with the dullest external lives; etc.