Originally Posted by Efthreeoh
Not to throw cold water on this whole statistics lesson, but I think the better data to make a statistical analysis from is time (rather than mileage) and type of use (i.e. percent split between city and hwy driving). The water pump fails due to the heat load it sees based on the environment it operates in. A water pump in car that sees more stop and go driving in Dallas, Texas will fail at a different mileage than a water pump in a car that sees no heavy traffic in Anchorage, Alaska. Tying failure rates to mileage, while you can make a statistical evaluation, is not a true measurement of predictability when a pump will fail as a result of mileage.
Also, throwing in N52s into the data set is going to skew the results because the water pumps for each engine are different parts. You should not be trying to correlate failure rates for two different models of water pumps.
Considering the high cost of replacement, it would be a disservice to members of this Forum to have them use this thread to decide when a preemptive replacement of the water pump should be made (i.e at 70,000 miles). Being that the cost for replacement is anywhere between $800 - $1,200 (non-DIY), it is probably better for an owner to wait until the pump fails and have roadside assistance insurance to handle the tow bill. A much better investment would be for an owner to buy a BMW-reading scan tool and periodically scan the car for the shadow codes that indicate the pump is starting to fail. Having been through this whole ordeal, this is just my opinion.
Certainly a valid point (The time metric is much harder to define), and I didn't intend for this to spur preemptive replacement. But it is good to understand when that it actually represents a reasonably high probability of failure. I added your quote to the OP...