Matt Yglesias wrote a piece mathematically apologizing for failed security efforts. In short, he claims that given Britain's population size, a 99.9% method of detecting terrorists would yield 1500 suspects, whilst he thinks only 15 actually exist. Given the huge number of false positives, identifying the actual 15 becomes very difficult.
He overlooks the monitoring capabilities of the Brits (or maybe just has a very realist appreciation of their abilities ;). With a 99.9% accurate model, one would rejoice and happily put the 1500 people under surveillance (aperiodically trail them for a day; monitor communications, contacts; always search them at airports; etc.). I'd probably be a cheeky bastard and provided a model of a good citizen, I'd tell a test subset of the suspects that they are on the airport search list and that to get off the search list they should do any combination of the good citizen attributes (married with kids, start a company, etc.) and see whether that changes their communications and contacts.