The Metrics Trap. . . And How to Avoid It
- 05 May, 2006 10:58
Calculating IT spending as a percentage of overall revenue is easy, popular . . . and misleading. There's gotta be a better way. In fact, there are several.
- Why calculating IT spending as a percentage of overall revenue doesn't distinguish between strategic and nonstrategic spending
- Where the wild variables live
- The pros and cons of popular IT spending metrics
Joe Drouin caught a lucky break when he became VP and CIO of TRW Automotive in 2002 - or so he thought at the time. A consultancy brought in to benchmark all of TRW's internal functions - everything from IT to legal to sales - found that the company was spending less on IT as a percentage of overall revenue than the industry average, which was about 1.5 to 2 percent.
Not one to look a gift horse in the mouth, Drouin played the metric for everything it was worth, highlighting it in every PowerPoint presentation he could during his first year as CIO. "I used it to say, we are managing IT effectively, and here's the confirmation from this outside firm," recalls Drouin. "It made me one of the good guys in the eyes of the CEO and COO." At one point, the CEO, who believed that inexpensive IT was good IT, joked that he expected to see Drouin and his staff outfitted with T-shirts that had the percentage stamped across their chests in big, block numbers.
But as that first year wore on, Drouin felt less and less like wearing that T-shirt. "I was guilty of using the number not because it demonstrated the value of IT but because it showed a positive trend," he admits.
The shallowness of the benchmark became clear as Drouin prepared his first budget presentation. The CEO asked him to break out IT spending as a percentage of revenue for TRW Automotive's 12 individual business units, each of which had its own legacy infrastructure and independent IT spending patterns. As it turned out, costs ranged significantly across the units. In fact, some units were spending two to three times more than others on a percentage of sales basis.
Suddenly, Drouin didn't look like one of the good guys any more.
He looked like a manager whose costs were out of control.
Much to Drouin's chagrin, the CEO initially tried to use the percentages to reduce the budgets of the higher-spending units. But it quickly became clear that the spending had no correlation to business success. Some of the units spending less on IT as a percentage of revenue were not doing as well as units spending more. Worse, if TRW Automotive's overall revenue fell (which was a distinct possibility given the auto industry's struggles), Drouin's IT spending as a percentage of revenue was going to rise even if he didn't spend an extra dime on IT.
"It's disappointing to be measured on one simple metric, only half of which I have any real control over," a sadder but wiser Drouin says now.
Caught in the Metrics Trap
Using percentage of revenue as a foundational metric to measure IT tends to cast IT in the role of a cost to be controlled. "When CEOs focus on spending as a percentage of revenue, it's because they've already decided to cut IT spending and they use the data as a justification," warns Barbara Gomolski, research vice president for Gartner. "It's hard to use this metric to show that you are doing the right thing." In this zero sum game, success is defined simply as lowering the percentage over time. "It's not clear how low it should go," says Drouin. "Joking with the CEO, I said: 'In your mind it should be zero.' We had a good laugh, but at what point do we decide it's at the right level and you don't drive it down further?"
Drouin wanted to burn those old PowerPoint slides. "I used the metric to my advantage, and it turned around and bit me," he laments.
It was too late to put the cat back in the bag. Nor was it an option to complain about how unfair the metric was. Rather, Drouin knew he needed to educate the CEO and COO about what constituted constructive cost-cutting in IT versus that which would rob IT of its ability to provide strategic differentiation. "I still mention the metric," says Drouin, "but I don't dwell on it." Instead, he emphasizes benchmarks that are more specifically targeted - such as the per-user cost of ERP systems in the different business units - to show that he's lowering costs in a way that will help the business rather than paralyze it.
But it's hard to get away from such a deceptively simple measurement - IT spending as a percentage of revenue - which remains "far and away the most popular metric for benchmarking IT spending", says Gartner's Gomolski, who's not at all happy about it.
"Top business executives don't want a dozen different metrics," Drouin explains. "The COO said to me: 'Come up with another metric, and I will leave this one alone.' I consulted the consultants and the analysts, and they couldn't come up with one."
The reason Drouin's consultants couldn't offer him a magic metric is because one doesn't exist.
There is no single metric for determining the right amount of money, the right percentage of revenue, to spend on IT. Even those CIOs who believe that measuring IT spending as a percentage of revenue produces a legitimate benchmark, such as Dow Chemical's David Kepler, consider it a very rough start to a much deeper discussion of IT investment strategy. "My experience is it's relevant to describe about how much you should spend," Kepler says. But, he adds, "it says nothing about how effective your spending is".
The problem most CIOs have to confront is that, in the view of much of the business, IT spending is a big black box. In fact, it's often the company's single largest capital expense and one that promises to grow larger as the world grows ever more digital. "The reason people attack IT over this metric is because they don't appreciate what you're spending and why," says Kepler. Unless CIOs can fill this gap in understanding with a delicate combination of relationship building, education and a series of contextual metrics designed to give businesspeople a sense of IT spending effectiveness, efficiency and value over time, CIOs will see their budgets reduced to a cost to be managed against a vague, often completely irrelevant, constantly sinking average.
Where the Wild Variables Live
The idea of describing and analyzing IT spending in relation to overall revenue has been around for a long time - decades, say some analysts. One can attribute its longevity as a business metric to its simplicity and the ease with which one can make the calculation. It's also easy to get the benchmarking information out of your competitors. No one's giving away any trade secrets by filling out (often anonymously) a survey with their IT spending information.
But it's also a legacy from an earlier era. "The metric is from a time when people viewed IT as a utility rather than as something to provide strategic differentiation," says Scott Holland, senior business adviser for The Hackett Group.
It was, in many ways, a simpler time. In the days when mainframes ruled the earth, IT was more centralized than it is today and, consequently, costs were much easier to calculate. IT back then was basically a handful of really big machines, and the software and staff needed to run them. By comparison, today's Internet and PC-based computing infrastructures are much harder to account for. It's pretty easy for a rogue marketing department, for example, to buy its own servers and software without IT ever knowing. It's much harder to sneak a mainframe past building security. And IT today is a challenge to pigeonhole. For example, should a mobile phone/PDA combination count as IT or telecomms? And does telecomms - traditionally separate from the IT budget - count toward IT spending? Is software depreciated over time or counted as a onetime expense? In all these cases, it depends on the analyst or consultancy doing the counting.
There are so many ways to account for IT that most CIOs dismiss the percentage-of-revenue metric unless it is carefully targeted for their specific industry or industry segment. Yet even in an industry context, the averages can mask huge variations in the sample. Companies at the high end can spend as much as 100 times more than those at the low end, according to CSC Consulting. "If I say that the average age of my two children is 10, do I have an 11-year-old and a 9-year-old, or does my family consist of an 18-year-old and a 2-year-old? It is not enough to know the average. One must also understand the spread of the data," wrote Eugene Lukac, a partner with CSC, in a report about a recent CSC spending survey. Nor does the average account for a company's unique situation; for example, costs will vary wildly between a chemical company with 10 plants spread around the globe and one with three plants, all in New Jersey.
The Battle Over the Numbers
Douglas Novo, former CIO for Venezuelan chemical company Polinter, knows all about these variables. The chemical industry is capital-intensive, dependent on expensive plants and equipment, with low profit margins and unpredictable revenue. That volatile combination means chemical executives are always looking for ways to cut operating costs. To do that, they bring in consultants about every two years to compare their company with its peers and to run the numbers.
In Novo's case, the consultants told him in 1995 that he was spending 1.5 percent of Polinter's revenue, or about 1 percent more than his peers. Novo's response was simple and blunt: "Impossible!" (Indeed, other sources told him that the industry average for the petrochemical sector at that time was about 2 percent.)
What was he to do? The numbers he was being shown didn't jibe with his notion of reality but neither did they come out of thin air.
"There's no equivalent of Generally Accepted Accounting Practices for IT spending," says William Mougayar, VP and service director for Aberdeen Group. "Without some kind of standard for how to account for spending, you can't use percentage of revenue to compare across individual companies. It's too hard to know what's being included and what isn't."
But now Novo was on the defensive. He had to justify $US2 million in spending that suddenly seemed to place him on the wrong side of the chemical industry norm. First, he pushed back against the consultants' findings. He got them to accept the idea that perhaps Polinter only seemed to be spending more than its peers because those companies were failing to take all IT operational and investment costs into consideration. Novo argued that the greater centralization of Polinter's IT operations, as opposed to its competitors, made Polinter's accounting more accurate. And why should he be penalized for his accuracy? But basically it was still Novo's word against the consultants.
So he fought back in a more constructive way: He created his own metrics. He took Polinter's history of IT spending and broke it down as a percentage of revenue, per employee, per employee served by IT, per IT employee and per ton of finished chemical product (a common metric in the industry). "Using historical data is important," Novo says, "because then you can show executives how spending has changed over time and how it responds to changes in the business, such as during bad years when revenues fall."
Novo's goal was to get off the defensive and shift the discussion from an argument over numbers to an analysis of performance. "It's important to focus on what you are doing - not someone else," he says. "The CEO wants to understand why you're doing what you're doing and what you're doing to improve."
Yet hanging over his discussions with his boss was the suspicion that by offering his own numbers Novo was simply making excuses to avoid having to cut his budget. To allay those suspicions, he divided his numbers into two pots: IT spending on ongoing operations (essentially the infrastructure and the services needed to keep the lights on) and IT spending on new applications, research and services. By slicing the spending this way, Novo could focus cost-cutting scrutiny on those parts of his budget that he can and should be cutting, while preserving the portion of spending that, if cut, could threaten IT's ability to provide competitive differentiation.
"The smart IT people are continually tuning the infrastructure to make it lower cost and more efficient so they can free up more funds for new work and innovation," says Laurie Orlov, vice president and research director for Forrester Research.
On Toward Better Metrics
The same thinking as Novo's guided Drouin to begin giving his COO and CEO data on the cost of providing employees in TRW Automotive's different business units with ERP software. The idea, says Drouin, is to cut the costs of providing important capabilities to the business without endangering the capabilities themselves.
"We want to drill into more granular kinds of metrics so that we not only get a real apples-to-apples comparison but also so that we find areas that are really worth investigating," Drouin says. Like Novo, Drouin also separates out IT operations spending, but his accounting task was much more difficult than Novo's, who runs a highly centralized IT operation in a single, homogenous business. But Drouin says tracking costs on IT operations across all the business units was worth it. "We could show that the old legacy systems and infrastructures were costing us much more in some units than in others," he says. "It helped us drive our business case for buying a single ERP platform to serve all the units. The project caused us to raise our spending, but we could show that it would lower the cost of providing ERP over the longer term."
Another increasingly popular way of slicing up IT spending for CEO consumption is by employee. "I do it that way because it mirrors the way the business consumes my services, so it's something they can relate to," says Novo. Yet this metric also has its drawbacks. For example, in the chemical industry, where capital costs and levels of automation are high, the per-employee cost of IT will be high ($US17,671 per employee, according to Forrester) compared with other, less capital-intensive manufacturing categories like industrial products ($US4302 per employee). The numbers will also be out of whack if the company has a lot of employees who never touch a computer (making the cost artificially low for employees who actually do use a computer) or if the company has outsourced much of its operations (which will raise the per-employee cost to a misleadingly high level because the outsourcer's employees will not be included in total IT spending). However, using the per employee metric in combination with percentage of revenue can help determine whether two companies in the same industry really are comparable - if, for example, both spend the same on IT as a percentage of revenue but have wildly different per-employee spending levels, chances are the business models, size, complexity or geographic spread of the companies are so different as not to be worth comparing, according to Gartner.
The accuracy of the total employees metric can be increased by slicing it up into actual users of IT, as Novo and other CIOs have done, eliminating from their calculations employees who don't consume any of IT's services. Hackett's Holland takes this a step further, taking into account nonhuman "users" in highly automated industries, like banks, where the per-unit spending on ATMs, for example, is every bit as important as the spending on a human teller or bank manager.
Talking Metrics with the Boss
The downside of these more precise accountings of IT spending is that they get away from the simplicity of the percentage-of-revenue benchmark that appeals to business executives. This is yet another reason why CIOs need to report to the CEO - or at least have a seat at the executive table. Springing these more nuanced metrics on the executive committee at budget time once a year just won't work. This kind of discussion needs to take place over time and, more importantly, in the context of the business's current strategy. For example, when Polinter went on an acquisition binge during the 90s, Novo added a per-acquisition breakdown to his spending metrics and began tracking how long it took to absorb a new business. He says that he was able to cut that time from 12 months to three months over four years. With regular access to the CEO (to whom Novo reported directly), Novo was able to align his IT spending - and cost cutting - to the company's acquisition strategy and justify the investments that helped facilitate it.
The more disconnected the CEO is from IT, the more likely he or she is to grasp at arbitrary averages to place the IT budget in context. And no CIO wants to be wholly judged by averages because that tends to make them and their departments average. For example, according to CSC's survey, companies that spend much less on IT than the average for their industry are three times more successful than those in the middle.
But companies that spend much more than the average are six times more successful.
A survey by Aberdeen Group found a correlation between poorly performing IT groups and their reliance on spending as a percentage of revenue as their primary benchmark: 62 percent of the poor performers used it, while only 5 percent of top performers did. Furthermore, IT departments that are better connected with the CEO, the company strategy and the business have a greater ability to cut costs and enforce standards, according to Hackett's Holland. Without that connection, and without enforced standards, companies will have needlessly complex IT infrastructures, thanks to rogue spending on IT by the business units. "The companies with the complexity are the ones where the CIO is an order-taker," Holland says, "because those companies can't control purchasing decisions."
Part of the reason that spending as a percentage of revenue survives is because it describes IT spending in simple, familiar business terms. And for CEOs and CFOs, business terms, even overly simplified ones, are the way they think about and run the business.
"IT executives aren't good at describing IT work in business terms," says Forrester's Orlov. "They describe it in terms of technologies they are supporting. IT spending should be described in terms of growing revenue, lowering cost and improving the time it takes to do something. If all you talk about is uptime, you are a cost centre, not a strategic partner."
"In the end, the real metric is not IT, it is business performance," says Dow's Kepler. "What is the output per employee? How efficient are we as a business per employee?
"If you're looking for a metric to justify IT spending, that's not the right mind-set," he adds. "The right mind-set is to understand how processes, systems and people tie together to get business results."
SIDEBAR: Your Guide to Popular IT Spending Metrics
AS A PERCENTAGE OF REVENUE
Pros: Easy to calculate; provides some basic insight into spending levels in specific industries. Widely used.
Cons: Averages can often mislead, masking wild variability in the sample. Fluctuating revenue can affect percentages. Doesn't distinguish between strategic and nonstrategic spending; doesn't describe impact, only costs.
LIGHTS-ON OPERATIONS AS A PERCENTAGE OF REVENUE
Pros: Distinguishes between strategic and nonstrategic spending; can show trends in improved IT efficiency over time, even if overall spending doesn't change.
Cons: More difficult to calculate. Definitions of strategic and nonstrategic IT vary across companies and industries. Not widely used.
Pros: Easy to calculate and widely used. Maps IT spending directly to those consuming IT services. Can be used with percentage of revenue to gauge the true similarity of companies in an industry.
Cons: Outsourcing can skew numbers. Numbers can be unnaturally high in capital-intensive industries. Does not describe IT effectiveness, only costs.
Pros: Can be useful for benchmarking the cost of well-understood categories of products and services, such as PCs or enterprise applications. Can be extrapolated into "unit cost" of automated services, such as ATMs in banking.
Cons: Does not differentiate between strategic and nonstrategic spending.
INDUSTRY STANDARD METRICS
Pros: Many industries have a favourite benchmark (such as "cost-per-ton" in the chemical industry) that is accurate and widely understood. Puts IT into a business context.
Cons: IT is generally one of the highest-cost services in a business. May make IT look egregiously expensive.
SIDEBAR: Why You're Spending More
If your business benchmarks IT spending as a percentage of revenue, these 12 variables may be making you look bad
1.Your product A high IT element in products (insurance and financial products, for example) usually means higher IT costs.
2.Business volatility If you are absorbing another company in an acquisition, your spending will be higher than normal.
3.Organizational structure A decentralized IT structure is more expensive to operate than a centralized one.
4.Competitive pressure If your company focuses on using IT to add new business capabilities, your spending will be higher.
5.Geography A global enterprise's IT is more expensive than that of a local or regional company.
6.Size Smaller companies tend to spend more on IT as a percentage of revenue than larger companies in their industry. (Big companies tend to spend more per employee because they tend to be more complex.)
7.Complexity Highly complex IT infrastructures (many different systems, many older systems) cost up to 50 percent more to maintain than low-complexity infrastructures (standardized, few applications).
8.Appetite for risk Aggressive adopters of new technology may outspend mainstream adopters by 30 percent and risk-averse companies by 50 percent.
9.Service levels High service levels (for mission-critical IT) cost more to maintain than low service levels.
10.Blip spending Upgrading a system or refreshing all the old PCs across the company can raise costs temporarily.
11.Rogue spending Companies where IT is decentralized and business functions have the power to buy their own IT will spend more.
12.Revenue per employee Companies with high revenue per employee will tend to have more knowledge workers who use IT intensively, thus pushing up IT spending levels per employee.
Sources: Aberdeen Group, Forrester, Gartner