Revolutive IT™
 
IT Metrics For Success

When done right, project-management measurement can boost a company's productivity

By Deborah Asbrand (http://www.informationweek.com/)

Despite years of coaxing, IT organizations are largely reluctant to institute project-management metrics programs. IT departments at small and midsize companies in particular seem disinclined to institute policies to track the performance of development projects. With limited management resources available and the pressure to quickly deploy new business solutions, productivity measurement often becomes a low priority for many organizations.

But some companies are bucking this trend--and reaping the benefits of improved productivity.

Belk Inc., a national retailer, was forced to adopt productivity metrics as a means of staving off devastating system failures. Conda Lashley, the veteran IT consultant that Belk hired, was used to nursing client organizations through crashes that periodically downed their systems. But nothing could prepare Lashley for the failure rate at Belk. Soon after joining the company as senior VP for systems development, Lashley discovered that Belk's batch systems went down an astounding 800 times a month.

The Charlotte, N.C., outfit, a private company with estimated annual revenue of $1.7 billion, paid a heavy price for the constant bandaging: In 1997, Belk spent $1.1 million of its $30 million IT budget on unplanned maintenance.

To steady the systems, Lashley instituted a series of tracking measures. Programmers began logging their time. Software function points were carefully counted in application development projects. Belk compared its cycle time, defect rates, and productivity with competitors' figures. And systems managers were required to draw up blueprints for reducing the crashes--with the results reviewed in their performance evaluations.

Costs Under Control
The transition to tracking the IT department's performance was painful but worthwhile, Lashley says. Belk's systems are becoming more stable--monthly disruptions are down to 480 incidents, a figure Lashley hopes to slash by another 30% this year. Unplanned maintenance costs also have been brought under control. Belk has cut unplanned maintenance expenses by $800,000 so far this year.

Among IT organizations, however, Belk's is the exception. Stern warnings and parental cajoling from consultants and academics on the importance of productivity measurements have been largely ignored for years by many organizations' IT executives. The adoption rates for productivity measurement tools and procedures are bleak. The perceived ease of object-oriented application design and the Wild West atmosphere of Internet development further discourage the discipline that the use of metrics requires, say observers. Still, experts hold out hope that more organizations will get the message about the importance of tracking performance metrics.

The advocates of application development metrics have their work cut out for them. In the United States, fewer than 9% of companies use metrics to measure and monitor software development, according to a 1997 poll of 1,100 companies by market research firm Rubin Systems Inc. That's not surprising given metrics' dismal ranking among technology priorities. On a list of 19 issues that included recruitment, productivity, and project management, metrics rated dead last.

What's more, three out of four measurement programs fail, according to research by the Yankee Group, translating into 1.5% to 3.7% of IT expenditures being wasted. Many IT management teams lack either the experience or the will to ensure the success of a performance tracking system. IT shops that succeed at measuring development output often bring some outside consulting expertise onboard to get things rolling. Once the initial implementation is under way, the consultants transfer knowledge to the IT manager so that he or she can use the metrics in future projects.

Why aren't more shops successful? "Application development is poorly managed--period," says Alan Gonchar, president of Compass America, a business-performance consulting group. Most organizations do a poor job of time reporting, he says. Even basic metrics--such as how and where programmers spend their time, and how many lines of code the organization maintains--aren't done, he adds. One of the first steps of year 2000 remediation, for example, is to figure out how much of the system must be fixed. "If application development was properly managed in the first place, you would already know that," Gonchar says.

Indeed, metrics make a lot of sense. In other parts of a company, the use of various performance yardsticks is routine. Sales representatives are compensated based on their ability to meet quotas. Marketing departments and other units that service internal customers regularly track their billable hours in an effort to account for their time and gauge their productivity. Profit and loss statements serve as universal benchmarks for overall company performance.

Air Of Mystery
With failures so common, why is there such stubborn resistance to applying a few measurements to application development, which typically consumes 10% of overall IT budgets? For starters, software development has long had an air of mystery about it. Software developers are often viewed as creative types who should be left to their own devices. Many developers who pride themselves on creating elegant technical solutions chafe at the notion of measuring their output.

Application development has historically bypassed benchmarking efforts because it has been perceived as a back-office function, particularly in large companies. "The key thing that has been lacking in the software development arena is that it hasn't been looked at as a necessary asset," says Dennis Huber, who, as VP of business information and technology solutions for Sprint, oversees 2,200 programmers and contractors.

That fundamental lack of understanding of where development fits into the corporate structure and how best to manage it helps explain the cost-cutting roller coaster that IT departments have experienced over the years. As the economy dips and rises, so do budgets. When senior management cuts IT budgets, efforts in time reporting, evaluations, and data-gathering--all of which should lead to greater efficiencies--are among the first areas to be cut.

It's an IT tradition that strikes many as paradoxical. "If I were a CFO, the first thing I'd put in place would be a measurement system," says Malcolm Slovin, VP and service director for performance engineering and measurement strategies at Meta Group Inc., an IT advisory firm. "I'd need to know inventory levels, backlogs, outflow, and what my competition is doing. The last thing I would get rid of is measurement. In IT, it's the opposite, which is a crazy thing when you think about it."

The explosion of the Internet and electronic commerce has only worsened the situation, convincing organizations that quick development times require less analysis of procedures. "In the age of the Internet, people don't think they need to do things" like measurement, says Howard Rubin, CEO of Rubin Systems, and chairman of the computer sciences department at Hunter College in New York. "There's a cowboy mentality that says you don't need this stuff to execute the work."

Metrics often seem like so much theory, worlds removed from the intense daily pressures and high stakes of large IT organizations. Advocates of measuring IT and software efforts say efforts to coax businesses into using IT metrics have been hurt by the academic nature of many techniques. Lines of code, function points, and information economics have all held the title as the newest and best barometer of IT productivity. Function points are features and functions that users request and receive in the applications the IT department builds. In 1998, balanced business scorecards and business value metrics are in vogue. These measurement theories track IT functions against the value chain of the organization.

No Single Best Method
The problem is that there isn't a single best method, and the various measurements offer little information without analysis. Several years ago, the Mutual Life Insurance Co. of New York plowed enormous resources into creating an inventory of function points from its 120 million lines of code. Was the effort useful? "Not really," says E.P. Rogers, CIO for MONY. "It wasn't telling us what we wanted to know. It gave us a lot of information. But there were such wide swings that it wasn't terribly meaningful." Someone could spend five hours on a job that had 100 function points, and someone else could spend 100 hours on a job with five function points, he adds. The trouble is that data alone doesn't tell IT organizations the information they need to know.

However, sticking with the effort shows substantial payback. The Software Engineering Institute's Capability Maturity Model estimates returns of four- or five-to-one for successful metrics programs. The CMM is a set of processes and procedures that assist development teams in progressing upward through five levels of quality and achievement. Its adoption is proceeding, albeit slowly, through large development teams.

Among the companies that have adopted CMM is the U.S. banking division at EDS. The division's returns using CMM are substantially higher than the 5:1 ratios seen across U.S. corporations, says chief technology officer Bill Wilkerson. The unit's 200 programmers are split into teams of 10 to 100 members, and there are two CMM Level 2-certified teams.

But convincing EDS clients of the value of CMM is another matter. The banks and insurers among the division's 30 clients have been slow to accept metrics. Of the 600 companies surveyed by the SEI in June 1997, only 3% were financial institutions.

Given a 20% return, the 5% investment of talent that CMM demands from most organizations seems a shrewd move, Wilkerson argues. He estimates that financial institutions require 35 software engineers to support every $1 billion in assets, and that annual overhead for each engineer is as much as $150,000 for taxable benefits, office space, and computing horsepower. For an institution that employs several thousand programmers, "if you can affect that by 20%, that's compelling," Wilkerson says.

MONY agrees. After dissolving its outsourcing agreement and bringing its IT operations back in-house last year, the company began to rethink its IT function and how it was using its 260 staff programmers. "We needed to know whether we were delivering what the business needed us to deliver," says CIO Rogers. The stakes are extraordinarily high: MONY's 1998 IT budget is $60 million, 50% higher than the insurer's usual figure. This year's allocations to IT have been increased to cover year 2000 fixes, as well as the costs associated with its plan to become a public company.

Rather than adhering to one measurement philosophy, MONY has combined several. MONY still tracks lines of code and function points on larger projects, but it has stopped using function points for code maintenance and support. It has implemented a balanced business scorecard, a management framework that relies on a broad range of indicators--customer perspective, internal processes, learning and growth, and financials--to reveal whether the organization is moving toward its strategic goals.

"We've focused on percentage of projects completed on schedule and on budget, defect rates, and customer satisfaction--issues that the customer is more worried about than internal IT is," says Rogers. "Our percentage of projects delivered on time has gone way up."

Sprint, too, has adopted an amalgam of function points and balanced-scorecard methods to measure its software efforts. The mixture is critical to success in a market as competitive as long-distance services, says VP Huber. But so is the reality check that Sprint's software effort gets by focusing on five interrelated areas: financials, process metrics, employee metrics, customer metrics, and leadership.

Pursuing just one area could hamstring development, says Huber. For example, speed to market is a critical measure. "But if I solve the problem for the marketplace and meet the customer's date and in the process work people 99 hours a week, then I destroy morale," he says.

Measuring Objects
Object-oriented and component-based design holds out tantalizing benefits for development shops--eventually. But as pro- grammers adjust to the impact of reusable code, organizations want to measure whether they're getting any real value from the new technology. The newness of objects and components is precisely the reason for trying to measure the technology's efficiency and effectiveness, say metrics experts. Objects "don't eliminate the need for measurement," says Chris Kemerer, a member of the information systems faculty at the University of Pittsburgh. "It's not magic." In fact, Kemerer says organizations' need for evaluation of object-oriented design is a major driver behind the adoption of metrics programs by many companies.

Retailer Belk will decide by November whether it will make a large-scale investment in object-oriented design. This summer, the company is evaluating the feasibility of harvesting some of the complex software controls now used in its legacy systems--some of which are two decades old--for use in component libraries.

Senior VP Lashley expects reusable software to improve the quality of Belk's systems because the programs have already been tested. He also expects reusable software to allow programmers to deliver new applications more quickly. "Why build something new when we already know this works and has been tested?" he asks.

In software, as in all things, attitude is everything. Part of the reason Huber's efforts at Sprint have been so successful is because he sees software development as essential to Sprint's business, and he conveys that to top management. "We're a phone company, but what it really comes down to is we're a software company," he says. "Everything we do is based on software. We're in one of the most competitive industries out there. For us to be successful, we have to have application-development metrics."

One of Huber's credos could be a mantra for all metrics advocates: "If you keep doing what you're doing, you're going to keep getting what you're getting."


Our solutions are the most efficient, flexible, quality-based and cheaper of the market, due to our fully innovative and customer-driven approach. Discover them and compare them to the solutions provided you by our competitors.


 
Copyright 2002 Atlantis Technologies. All rights reserved.