The IP Watchdog has a nice entry today about patent searches.
Some of the points that he makes are nothing new, but I've never before thought of or heard of them applied to patent searches, such as the idea that as the certainty of the result increases, the expense in getting there does too. This is a standard idea from statistics, but then this leads to some interesting outcomes. If you are just applying for a patent, there is not as strong of a need to be 100% certain that there is no prior art as that would be too expensive at the application stage. At later stages (i.e., entering production), the need to reduce risk increases, so there is stronger justification for spending more on a more detailed search.
One line from the blog may raise your eyebrows: "In patent litigation it is not uncommon for defendants to spend upwards of $1 million looking for prior art, sometimes substantially more." The million dollar figure would only apply to the biggest of infringement cases, but considering some of the sums that Article One Partners is willing to lay out for just one good bit of prior art, I can easily see a million dollars being spent.
In this day of trivial internet searching, it is easy to get complacent and assume that everything of importance is on line. Book are a huge exception, although that is slowly changing, but the difficulty is that they have to be manually searched (after being manually located!).
A patent examiner is a lone individual looking for prior art, so it can be safely assumed that they will occasionally miss something. But when a patent is issued, the whole world can suddenly be looking for prior art, sometimes forcing the whole novelty story on a patent to quickly change.
Getting a patent is easy, keeping it is difficult.