Problems can often be classified as either leaky bucket or power law problems. A leaky bucket requires you to fill all the holes before it will hold water. By extension, a leaky bucket problem requires a solution that addresses all the aspects of the problem holistically, or else nothing meaningful is accomplished.
By contrast, a power law problem exhibits the Pareto principle, where you can get 80% of the effects (holding water) by focusing on just 20% of the caues (holes). Power law solutions are all about focusing on one piece the ecosystem that happens to get most of what's meaningful accomplished.
As a startup you want to address power law problems when possible because they require less dominoes to fall perfectly in place before you can succeed. However, it's often hard to determine what you have, i.e. whether you really have a leaky bucket problem on your hand or not.
For example, I'm struggling with that question right now with regards to the Philly startup scene. A lot of the compelling advice out there says you need a fully functioning ecosystem to create a truly great startup hub. That's classic leaky bucket thinking with lots of moving parts and holes to plug. But is that right?
My gut tells me that it is really a power law problem and you could focus hard on one thing and accelerate the whole process. I have that inclination because fundamentally it doesn't take that many people to start a bunch of ambitious and successful companies, and so there should be a way to focus attention on those people and let the rest fall into place later.
Yet it isn't clear what exact solution to implement, and that's generally the issue startup founders face pre-product/market fit. You see all these holes and you naturally want to plug them (after all, they're leaking stuff!), but usually doing just one really well is the right path. That's why I'm generally a big fan of the strategy to make educated guesses, try a few things a little bit and then focus hard on the one that is promising. If it pans out, keep at it. If it doesn't seem like it is going to exhibit the Pareto principle, then back up and try something else. In other words, be a heat seeking missile.
This power law thinking is tough because we aren't wired to think that way. But the trap is to get caught out spinning your wheels addressing all that product feedback and missing the golden nugget hidden inside it.
Nevertheless, there are real leaky bucket problems. A search engine is a canonical example. There are so many edge cases and no one really wants to use you as their primary search engine until you've covered a lot of them, and you still lose people every day because of them. That's the story of DuckDuckGo, my current startup.
At the same time, I constantly wonder if that's really the case and there isn't something else that can cut through this leaky bucket problem and turn it into a power law problem.
A definitive leaky bucket problem is a real viral loop, like I had in my last company. If your viral coefficient hovers around one then you need to fix every issue to keep it there since it is the difference of exponential growth and none.
Incidentally, a viral loop is a leaky bucket problem embedded in a Pareto problem: traction. Usually one traction vertical dominates when you exhibit step function growth and really move the needle.
 I realize the metaphor doesn't totally work with the picture in that you can plug a few and still hold some water.
 Since you can't have exponential growth for that long, all sorts of leaks spring that force you back down (under k=1) and it is a constant battle to close all these holes.