Revenue per employee is still being used to measure the productivity of information resources. The most notable recent example is the publicity given to a Harvard Business School study. The study is an example of how misleading measures lead to deceptive findings.

The study

Researchers chose insurance companies because they spend a higher share of operating expenses on computers than most economic sectors do. The findings were surprising. It was the firms' approach to downsizing - and not IT - that made the difference in productivity. It was not technology, but the behavioral and organizational characteristics that mattered.

The implications of those findings are readily apparent: It's not what you do, but how you do it that delivers the results. Unfortunately, the problem with these conclusions is that they all hinge on productivity as defined in terms of revenue per employee. That measure is unreliable. Therefore, although I agree with some of the observations, the research conclusions aren't trustworthy. Here are the reasons:

1. Head count doesn't account for the cost of doing business in insurance. The average salaries in insurance firms vary greatly, and therefore, every head is not equal. The total cost of operations must cover payroll expenses, as well as computers (up to 30% of payroll dollars), telecommunications, purchased services, office facilities, outsourcing fees and capital charges.

2. The Harvard study used life insurance premiums as the measure of output. But insurance companies have diversified and have other revenue streams besides premium income. These include fees such as custodial charges and income from investments.

3. If output (defined as premium income) and input (defined as head count) aren't reliable measures, then productivity (the ratio of output over input) won't be reliable. Revenue per employee wouldn't relate to any claims of productivity gains, particularly regarding IT, since technology use affects operating costs as well as profit margins.

Why head count-based measures are wrong

I'm sensitive to this distinction because in my years as a CIO in business and government, I witnessed the enormous damage done by thoughtless and arbitrary hiring freezes, forced retirements and downsizing. Head count ratio methods gave central planners a blunt and crude hatchet when thoughtful ways to boost the effectiveness of individuals could have delivered the identical results with less pain - and at a lower cost.

CIO Implications

Using statements filed with the Securities and Exchange Commission, I analyzed 34 insurers whose financial records stretched back to 1987. Nineteen added to their employment rolls in a 10-year period, while 15 downsized. I calculated the actual productivity numbers as a ratio of revenues from all sources, divided by the total operating costs. For those adding employees, the productivity improvement was 5.2% in 10 years. For those downsizing, the improvement was 6%.

Those numbers amount to average annual gains of only 0.5% - a poor showing when one considers that during the past century, the entire U.S. economy delivered average annual productivity gains of more than 2.5%. There was little difference between the productivity gains of those upsizing vs. those downsizing.

Whatever gains in profits shareholders accrued in the past decade are rooted in the steady decline in the costs of capital as interest rates deflated. The gains didn't come from improved workforce productivity, regardless of how many computers were installed.

When the steady decline in interest rates finally reverses, as has always happened before, improved profits will have to depend on productivity gains. CIOs should adopt realistic measures to track such improvements.

Their expectations deserve better counsel than that offered by the Harvard Business School researchers.