Big, Good things. Amazon S3 and Google Maps API are the only real ones that come to mind. You couldn't do either of those better if you tried.
It's the Big, mostly good things that are the most dangerous. Off the shelf Content Management Systems come to mind as the obvious example. They'll get you close fast, then morph into a soul-draining time sink as you try to get that last 5%. You'll know early on that you're on the wrong path, but it will likely be six months of torture and waste before you actually scrap it and build your own.
Perhaps when the opportunity cost of building is greater than the actual cost of purchasing. This frequently is true when specialization or large scale production makes the producer more efficient than someone who does not have those attributes.
For example, when it's time to eat you buy prepared chicken from a store; you don't spend a few hours every day tending a small coop of chickens and bring one in every week.
A counterpoint to this is companies that are funded with a high valuation. Investors expect to see a return on their investment and don't want to hear "we left the money in the bank while we try to find and hire a person that is capable of building something we could purchase". The seed/self funded case and overcapitalized case both have "have to build" == "get to build", but in one it's 1==1 and in the other it's 0==0.
When something is not core to what you're doing. Do the core stuff yourself, buy non-core stuff. The advantage of no cash: don't buy non-core stuff, but also don't do it.
How about when you're growing but trying to avoid making extra hires beyond the founders (or still trying to find the right people) and you can buy things to save some time that you're very short of.
When buying costs significantly less than building. To us geeks its inconceivable... "I can build one better" is a common phrase, but our managers keep us in check by saying "That is great, but through distributed cost THEY can build one better and cheaper!" which is true some of the time. It is usually our job to see if in fact we can build one better and cheaper or if that product is in fact superior. Geeks and Managers keep each other in check.
