``Science'' in the sense that many folks use the word almost never works.
Let's wax philosophical for a moment. If you're interested in scientific statements -- of which there are none below -- you might want to return to the chemical kinetics explanation now.
What do we mean by ``science?'' Perhaps the most widespread notion is that science is ``a well-defined, reliable method for discovering the answers to any questions you may have about the natural universe.'' This notion is false, perhaps dangerously so. Here's why: any nontrivial explanation of natural fact requires guessing the existence of underlying, simplifying principles that correlate and explain a flood of otherwise meaninglessly jumbled measured data.
Why do we need to guess? Because there is no way to figure out a principle just by observing its consequences. If we see water on the floor, we cannot simply by observing it know how it came to be there. Ice might have melted there, or a glass of water been spilled, or the dog might have forgotten his housetraining. If we also observe that it is raining outside, then we might guess that the water leaked through the roof, and then test our guess by searching the ceiling right above the water for a stain or drip.
The hardest part of explaining things is this necessary guessing of the possible underlying reasons. The genius of Sir Isaac Newton is not really so much that he was slick at math (although he was that, too), but that he could look at the complex patterns the eight naked-eye heavenly bodies scrawl nightly across the sky, and then guess that every twinkle of the ponderous ballet follows ineluctably from three to five simple statements, such as: ``an object moving in a straight line will continue to move in a straight line unless acted upon by an outside force.''
Guessing is also the first and most essential part of explaining things. How can you make any real progress until you have guessed a possible explanation? You can measure things as much as you want, but unless you have some idea that connects your measurements, you haven't really understood anything. (Let's also note that until you have thoroughly tested your idea for possible falsehood you still do not know anything. You merely have a supposition, a hypothesis.)
There is no way to automate the process of guessing -- to write a recipe or computer program for it -- simply because we do not know how to program creativity. Nor can we teach it. We can, in the classroom, go into great detail about how to test your guesses, study great guesses and guessers of the past, give you practice problems with which to polish your guessing skills, but in the end no one knows how to teach you to guess well. It is, without doubt, an art.
Because we cannot automate it, or teach it, or describe how it occurs, we also cannot predict how long it will take. How long will we sit, staring at the water on the floor, pondering, before we guess that the water might have leaked in from the roof? Half an hour? All day? Thirty seconds? Will it puzzle our grandchildren still fifty years hence? There is no way to know, no way to predict how long it will take before the golden, correct, beautiful idea, that makes all our observations ``click,'' blooms in the imagination.
Believing that you can automate or predict in detail the progress of guessing leads to such grievous follies as thinking that scientific progress can be planned, or that one can figure out where to invest research effort today so as to have the maximum possible payoff in terms of new technology fifteen years hence. (Incidentally if you are inclined to disagree with this, because you are a member of Congress or otherwise sheltered, you might ask yourself why Microsoft did not invent the Web browser nor IBM the personal computer. In both cases there was plenty of research money, no lack of intelligent management, copious in-house technical expertise in appropriate fields, and obvious [in retrospect!] motivation for doing so.)
What is predictable, reliable, and programmable about science then? In the narrowest sense the scientific method is just a well-defined, reliable set of rules for testing guesses about the explanations for natural phenomena. The rules include things like: ``try to predict something really general, so that a single simple experiment can prove you wrong, if you are wrong,'' or ``try to predict factors that should have no effect whatsoever on an experiment [mass of the Sun, number of toes on green macaque monkeys, which scientist is doing the experiment], so that if any effect from these factors is observed your guess can be immediately proven wrong.''
You'll notice that these rules are all about how to prove yourself wrong efficiently. There are no rules about how to prove yourself right, because that isn't considered a possibility in the scientific method. To paraphrase the great physicist Richard Feynmann: being scientific consists of doing everything you possibly can to prove your idea wrong. If you completely fail in this effort, then you may have discovered something true (or, of course, you may simply not have tried hard enough). The value of the scientific method is that it enables you to quickly prove your ideas wrong, so that you can move on to other, possibly more fruitful ideas.
In this rather narrow and gray sense science is the reliable, unimaginative and powerful machine it looms as in the pop-science press. It's also, of course, not what most people who do science spend their time doing.
But if you consider science to include the (fun) and time-consuming part, namely coming up with all the possible explanations for what's driving you nuts about the universe, then scientific progress is no longer linear and reliable, because the uncontrollable element of creativity has entered in a fundamental way.
In fact, it's a given that all but one of the possible explanations one can dream up for experimental facts are going to be wrong. Naturally, for any reasonably complicated set of facts, you're going to come up with a lot of wrong explanations before you hit the right one. Hence in the expanded view, where the most vital task of natural philosophy -- proposing new explanations -- is considered ``science,'' we must conclude, sadly, that most science doesn't work.