We routinely read elaborate tabulations of average IT-spending-to-corporate-revenue percentages by industry. For instance, last year's benchmark for the banking industry, according to Meta Group, was 5.16%; from Gartner, 4.1%; and from InformationWeek, 8%. In telecommunications, the numbers were 5.8% (Meta), 10.5% (Gartner) and 17% (InformationWeek). Similar disparities can be found for other industries.

If you have been asked to prove that your IT spending is reasonable, the best way to answer is to use the highest number you can find. Then you can plead for more money if your original budget request was less than what's supposed to be "normal" for your industry. Yet, one CIO used below-average spending levels (at performance appraisal time, not budget review time) to argue that spending less than "normal" was evidence of superior IT management.

The problem with an average is that the numbers depend entirely on what you measure and how you measure it. The damage occurs when irrelevant averages are used to justify risky decisions.

In that vein, I believe that IT/revenue benchmarks are irrelevant. But why have they been so popular? The most plausible answer is that they're easy to explain. But the real reason has to do with the changing structure of U.S. businesses. Rising revenues, especially during inflationary periods, place all IT spending on an automatic escalator if a firm sets spending levels using a fixed IT/revenue benchmark. As firms outsourced growing shares of their costs, revenues kept rising, even though payrolls were shrinking. This way, IT executives could rationalize more spending, even as the prices for IT products and services were falling.

The IT/revenue percentage is also a handy way of explaining that IT shouldn't be driven by economics, but propelled by an "arms race" to match whatever competitors are doing. To get around financial justifications, such as ROI calculations, all you need to do is find a consulting service that would demonstrate that your competitors have a higher average IT/revenue percentage.

Perhaps the best way of illustrating the irrelevancy of such industry-related metrics is to pick a relatively stable and technologically conservative industry sector: manufacturing. According to Gartner, the average IT/revenue percentage in manufacturing last year was 2%. According to Meta, it was 3.76%. So, I took a random sample of data from 74 manufacturers producing similar goods. My findings? Even though all the firms were grouped into the same industrial category, they displayed a remarkable range of financial metrics. Revenue per employee (an indicator of how much a company outsources functions) ranged from a high of $330,225 to a low of $45,496.

Transaction costs per cost of goods ['Tail' Over 'Teeth'?, April 2002], which indicates the amount of information processed to sell those goods, ranged from 97.2% to 3.1%. The sales, general and administrative expense per employee (which indicates the amount of IT needed to support revenue) ranged from $92,837 to $8,944. So, applying the IT/revenue percentage as a "normal" benchmark to all manufacturing firms wouldn't recognize their enormous differences and would be misleading.

Executive Implications

IT spending is not a characteristic of an industry, but a unique attribute of how a particular firm operates. Revenue-based "profiling" based on industrial classification is as questionable as other approaches based on criteria that don't recognize a company's specific characteristics. For instance, a superior steel company's numbers could look more like those of a superior computer manufacturer than those of an inferior steel company. CIOs should stop taking an easy (and easily manipulated) path for explaining their spending plans. Don't ask what others tell you is the "right" spending, but commit to what profits you can deliver for the company, whatever the cost.