I recently blogged about designing metrics, and the topic deserves a little more emphasis.
The point was that any time you give someone an incentive to hit a goal, you're also creating an incentive to manipulate the metric.
If it's easier to game the system than to actually meet the business goal, then you're going to have problems.
At our local Home Depot, for example, they've been running register tape surveys for as long as I can remember--you know, the surveys where they print a code on the receipt and ask you to go online to take a survey for some miniscule chance at winning a fabulous prize.
A couple years ago, a homemade sign appeared by the exit door with a picture of the cartoon "Home Depot Guy" urging customers to take the survey, helpfully informing us that "9 or 10 = PASS, 1 through 8 = FAIL."
As an attempt to bias the survey, this was pretty blatant. Most people don't like giving someone else a failing grade, so telling customers that only 9 or 10 qualifies as a passing grade will push the scores way up.
After 18 months or so the sign disappeared. I'm guessing that what happened is that this Home Depot store was failing to hit the internal goals for the customer satisfaction survey (I'll ignore all the methodological problems with register tape surveys). Corporate HQ came down hard on the store manager to improve survey scores Or Else.
Improving customer satisfaction can take some effort, thought, and possibly money. Then the manager realized that he didn't actually have to improve customer satisfaction: he just had to improve the survey scores.
So for ten bucks, an hour at Kinko's, and some cut-and-paste work, he made a couple of signs which would have the desired effect. The signs stayed until either a new manager saw things differently, or someone from HQ finally noticed and ordered them removed.
In the meanwhile, the manager met his survey goals, got his bonus, and nobody was the wiser for quite some time. All for far less effort than it would have taken to actually improve service.