If it makes you feel better, compensation has improved.

I’m not talking about the amount. I’m talking about how it’s administered.

Not that long ago, many companies gave men raises when they got married. Women’s careers stalled under the same circumstances.

The logic was impeccable: Married men needed more money to support their households. Married women, on the other hand, no longer needed as much. After all, they had husbands to support them.

The world does sometimes get better. Achieving perfection is another matter. If you expect it, you’re sure to be disappointed, especially where compensation is concerned.

This, and other thoughts, occurred to me as I read my correspondence from last week’s column on fair compensation (“Poor Joe,” Keep the Joint Running, 10/22/2007). What other thoughts? Glad you asked.

Why not base pay on value? Most employees think their pay is and should be based on the value they contribute.

It’s a reasonable thought. That value is why companies employ people in the first place (for more, see “The 70% solution,” IS Survival Guide, InfoWorld, 3/4/1996).

Here’s the problem: Whenever a company bases compensation on anything other than what the market will bear, it faces one of two situations. Either it can get the same value for less by replacing its employees with less expensive alternatives, or its employees will leave for better paying employers.

Basing pay on value is inherently unstable.

There’s another problem with value-based pay: Measuring the value. When you’re dealing with the sales force it’s easy. They sell. What they sell has a known margin. Do the math. No problem.

For just about anyone else, the connection between their work and the value they deliver can’t be turned into a number that will make much sense. The value is there. Measuring it unambiguously isn’t a problem any of us are likely to solve any time soon.

Can you really give a $75,000 employee a $20,000 annual bonus? No, you probably can’t. The problem is that every year you employ this computation, you’re assuming the employee will stay another ten years. Some employees will; many won’t. If that sort of longevity is typical in your company, annual bonuses this large might not be a bad idea. If it isn’t, don’t fret.

First of all, unless your company is a truly awful place to work, the average employee turnover in IT is unlikely to exceed 20%, which means the average duration of employment is at least five years. That still allows more than $10,000, which isn’t bad at all.

There’s another alternative, too, at least in publicly held or soon-to-be publicly held companies: stock options, vested over ten years. It’s true that the accounting for stock options has become controversial. The basic idea remains sound. Vesting options over ten years eliminates the risk of basing compensation on the assumption of ten years while getting much less, and in addition creates a financial incentive for employees to stay with the company.

In fact, you can give a smaller option grant and vest it over five years, especially if your company has a reasonable track record of stock price increases, because it’s the employee’s expectation of value that matters most.

Not a bad set of outcomes.

Pay for performance? Great. One question: How do you measure performance? This is the magic question, whether you recognize performance with variable compensation or prefer to use simple raises.

As a manager you have to have a way to assess … not measure, assess … how well employees are doing their jobs. The distinction between measurement and assessment is vital.

When you measure — when you establish clear, objective criteria regarding how well each employee accomplished the goals, objectives and responsibilities you’ve established — you face a hazard: You get what you measure. That means anything you mis-measure employees will get wrong, when you measure the wrong things you’ll get the wrong results, and anything you don’t measure you won’t get. Especially, what you don’t measure you won’t get.

Think of what you ask employees to do, all the time, in addition to their formal responsibilities: Participate in ad hoc committees; take the initiative when something needs doing that you don’t know about; and help each other out when one gets stuck and another knows the solution; to name just three typical examples.

If you assess employee performance, these will count in their favor. If, instead, you measure it, when you ask you’ll get a predictable, and entirely deserved response:

“Why would I want to do that?”