Since this is a hot subject on the software-testing mailing list I though I put some lines about it.
First of all I myself have little clue on how a good metric for testing should be. I know this causes a lot of problems in testing.
I am not saying not using a metric at all should be the solution because its impossible in an organization. There is the need for a formal one at least.
But let’s not focus extremely too much on it because let’s face it:

• Promotions are established well in advanced without any relation to real productivity, possible even before hiring
• Raises are given to all the employees usually and is percentage of salary most of the time
• Firing people is mostly done in cases of financial troubles

And there are many other situations when using metrics for employee evaluation that are completely un- useful.

Usually the general view is the one deciding and for the most part is the best. Enforcing metrics will increase the chances for sabotage among employees or give false results.

I don’t have a valuable solution for this but it would help differentiating testing metrics to development metrics. Because:

• Testing is considered a simpler activity for a specific functionality than developing it, but testing has to deal not only with a task-done type of thing, but larger areas, interactions, thing that shouldn’t be there. Is basically an infinite process
• Testing is more closer to marketing and arts or puzzle solving than computer science
• Testing is little bit trying to break the toy than building it up-so again common metric is not relevant

Things that a tester  does in addition to testing can be measured: installing an environment, reporting, presence at work. But is not relevant to the main job itself.

Also what should not be done (!!!) is rewarding testers by completely absurd criterias like:

• Times per week going to gym
• Playing guitar lessons
• Speaking loud and much
• Sabotaging skills
• Blaming skills to pay
• More bills
• etc

Although those above have some relevance it should not be considered a major factor because this means that testing in that organization is not taken seriously and doesn’t have a real chance to improve.

Metrics are indeed hard but if you use a group of sub-metrics and a good impartial general view you can tell what tester is more productive than the rest after some period.

It’s more like evaluating a person who is a better friend: is it the one who goes with you the most of the time to beer?  Could be – but is hardly a good criteria; it could be a  sub-metric though that can influence it.