Page 1

Agile Project Management Vol. 5, No. 7

Pushing the Envelope: Managing Very Large Projects by Ken Orr, Fellow, Cutter Business Technology Council

Very large software projects are being tackled in high-intensity environments, with heavy user involvement — and with a higher chance of failure. Why? This Executive Report answers this question and argues that despite widely held beliefs, agile techniques can be useful in taming these IT beasts.

Cutter Business Technology Council Rob Austin

Tom DeMarco

Access to the


Christine Davis

Lynne Ellyn

Jim Highsmith

Tim Lister

Ken Orr

Ed Yourdon

About Cutter Consortium Cutter Consortium’s mission is to foster the debate of, and dialogue on, the businesstechnology issues challenging enterprises today and to help organizations leverage IT for competitive advantage and business success. Cutter’s philosophy is that most of the issues managers face are complex enough to merit examination that goes beyond simple pronouncements. The Consortium takes a unique view of the business-technology landscape, looking beyond the one-dimensional “technology” fix approach so common today. We know there are no “silver bullets” in IT and that successful implementation and deployment of a technology is as crucial as the selection of that technology. To accomplish our mission, we have assembled the world’s preeminent IT consultants — a distinguished group of internationally recognized experts committed to delivering toplevel, critical, objective advice. Each of the Consortium’s nine practice areas features a team of Senior Consultants whose credentials are unmatched by any other service provider. This group of experts provides all the consulting, performs all the research and writing, develops and presents all the workshops, and fields all the inquiries from Cutter clients. This is what differentiates Cutter from other analyst and consulting firms and why we say Cutter gives you access to the experts. All of Cutter’s products and services are provided by today’s top thinkers in business and IT. Cutter’s clients tap into this brain trust and are the beneficiaries of the dialogue and debate our experts engage in at the annual Cutter Summit, in the pages of Cutter IT Journal, through the collaborative forecasting of the Cutter Business Technology Council, and in our many reports and advisories. Cutter Consortium’s menu of products and services can be customized to fit your organization’s budget. Most importantly, Cutter offers objectivity. Unlike so many information providers, the Consortium has no special ties to vendors and can therefore be completely forthright and critical. That’s why more than 5,300 global organizations rely on Cutter for the no-holds-barred advice they need to gain and to maintain a competitive edge — and for the peace of mind that comes with knowing they are relying on the best minds in the business for their information, insight, and guidance. For more information, contact Cutter Consortium at +1 781 648 8700 or

Pushing the Envelope: Managing Very Large Projects AGILE PROJECT MANAGEMENT ADVISORY SERVICE Executive Report, Vol. 5, No. 7

by Ken Orr, Fellow, Cutter Business Technology Council For large projects in the 10,000-100,000 function point range, the results are very grim indeed. More than half of these massive projects will be cancelled outright, and the percentage finishing approximately on schedule is less than 20%. This is a very troubling situation, because large systems are extremely expensive, and thus the failures can cost millions of dollars. — Capers Jones [1]

FOREWORD Serious engineering disciplines devote enormous amounts of time and effort to researching failures, documenting what happened and why. When a bridge, a building, or a walkway in a hotel collapses, careful analysis of the cause follows. The same is true

of transportation failures. In the US, when airplanes, trains, or ships collide with something, the National Transportation Safety Board teams are among the first to arrive on-site. In our domain, however, IT is not nearly as good at analyzing its failure. While IT should be good at this, it isn’t. One of the reasons that IT is not as good at analyzing its failures is that IT disasters are rarely as public as those concerning buildings or transportation. People rarely die (at least directly) as a result of IT failures, and most large organizations are not eager to air their dirty wash in public. With building and transportation failures, those involved have little choice. Once in a while, however, a blatant case of IT failure manages to

reach the front page. Recently, for example, the Royal Bank of Canada made the news because of problems created by a software upgrade.1 And a few years back, Foxmeyer Health Corporation, a distributor of pharmaceuticals and medical supplies, blamed its bankruptcy on poor implementation of a very large enterprise resource planning (ERP) project. But for the most part, IT failures, even major ones, largely go unreported. But very large IT failures continue to happen, and organizations are losing millions, perhaps billions, because of them. So if developing large software systems is to become more professionalized and more reliable, we must do a much better job of defining, architecting, designing, and building these very large systems. Based

2 on the public’s skepticism about the part of technology that people deal with every day, IT must do this soon, before software is considered too risky to attempt anything really big or new and the world loses one of its most powerful tools of innovation. INTRODUCTION From one point of view, this Executive Report is about scaling up agile, or incremental, development to the point at which it can be used successfully on very large projects (VLPs). In this context, the report serves as a discussion about how one combines environments characterized by high intensity, high user involvement, and uncertain requirements with the structure needed to complete VLPs. The report argues that VLPs have certain unique circumstances and that architecture plays a unique role. Many people believe — wrongly, in my opinion — that VLPs cannot avail themselves of agile techniques because, as the project grows, the problems of communication, coordination, and control become too difficult for an approach that depends heavily on face-to-face communication.

AGILE PROJECT MANAGEMENT ADVISORY SERVICE But beyond agile development, this report also focuses on something close to my heart: the cause of VLP failures. For years, industry experts have pointed out that IT project risk increases dramatically with project size. Unfortunately, the situation hasn’t gotten better. Not only do VLPs fail far more frequently than do small and medium-sized ones (see Table 1), but the rate of failure hasn’t changed much over the years. Despite the fact that hundreds of books and thousands of courses are offered on project management, large projects continue to fail on a kind of logarithmic scale. Despite everyone’s best efforts, large projects, particularly the very largest ones, continue to be extremely risky undertakings not only for the companies involved but also for all the individuals closely associated with it. A failed project doesn’t look good on anyone’s résumé. Not only do Capers Jones’s figures in Table 1 indicate that the number of projects delivered on time decreases by 40% as one moves from medium projects (with less than 1,000 function points, or FPs) to large projects (with more than 10,000 FPs), but the rate of cancellation more than doubles. And it

gets much worse for projects with greater than 100,000 FPs. The most obvious solution would seem to be to break VLPs into much smaller parts. But there are practical reasons why this can’t always be done. There are some very big projects that must be completed to meet very big business or organizational demands. But new approaches need to be applied that reduce the risks posed by VLPs and improve the odds of success. VLPs GONE BAD While IT does not have a resource that collects and studies failures on VLPs, one area provides some documentation on VLP failures: government systems. Because of the size, cost, importance, and constant scrutiny of large government activities, we probably know more about failures in this domain than in any other. For this report, I’ve included two notable cases to provide a baseline. The IRS Modernization Program

The IRS’s $8 billion modernization program, launched in 1999 to upgrade the agency’s IT infrastructure and more than 100 business applications, has stumbled badly. The first of multiple software releases planned for a new taxpayer database is nearly

The Agile Project Management Advisory Service Executive Report is published by the Cutter Consortium, 37 Broadway, Suite 1, Arlington, MA 02474-5552, USA. Client Services: Tel: +1 781 641 9876 or, within North America, +1 800 492 1650; Fax: +1 781 648 1950 or, within North America, +1 800 888 1816; E-mail:; Web site: Group Publisher: Kara Letourneau, E-mail: Managing Editor: Rick Saia, E-mail: Production Editor: Lauren S. Horwitz, E-mail: ISSN: 1536-2981. ©2004 by Cutter Consortium. All rights reserved. Unauthorized reproduction in any form, including photocopying, faxing, and image scanning, is against the law. Reprints make an excellent training tool. For more information about reprints and/or back issues of Cutter Consortium publications, call +1 781 648 8700 or e-mail

VOL. 5, NO. 7



three years late and $36.8 million over budget. Eight other major projects have missed deployment deadlines, and costs have ballooned by $200 million. This case study illustrates what can go wrong when a complex project overwhelms management capabilities of both vendor and client. — Elana Varon [11]

The cover of the 1 April 2004 issue of CIO reads: “IRS Files for Extension Again.” And an article by Elana Varon spells out, in excruciating detail, the IRS’s failure to bring its 1960s-era vintage Master File system into the 21st century. Launched in 1999, the entire project involved billions of dollars and various applications. High-profile management was brought in to manage the project, and a major outside contractor was chosen for implementation. But by the end of 2003, the project was classed as a failure, the contractor was constrained from bidding on certain IRS projects, and all participants had egg on their faces.

Table 1 — Distribution of Project Results by Size in Function Points (FPs) [1] Probability of Softw are Project Results Early

On Time



1 FP






10 FP






100 FP






1,000 FP






10,000 FP






100,000 FP












In the late 1970s, President [Jimmy] Carter put the kibosh on a project to replace the aging Master File when an external review questioned whether the agency could adequately protect taxpayer privacy. Almost two decades later, in 1995, Congress pulled the plug on a second modernization program after the IRS spent 10 years and $2 billion with little to show for it” [11].

as the culprit. Those involved lament: “If we only had the right vendor”; “If we only had the right contract”; “If we had people with more experience.” And yet, despite all this experience, all this management, and all this money, modernization at the IRS fails. Perhaps something else is wrong; maybe it is more than project management that the IRS lacks. The FBI Trilogy Program

As you might expect, for a project like this to make the trade press, things have to be pretty bad. But this was not the first time the IRS had failed in its attempt to build a better collection system. For at least the past 30 years, various attempts have been made with similar results.

Despite this series of events, a member and former chairman of the IRS Oversight Board told Congress, “People ask, Is modernization going to fail? I say we can’t let it fail”2 [11]. Those charged with overseeing the IRS continue to make comments like this, yet modernization at this government institution keeps failing. We should bear in mind the old saying, “Hope is not a strategy.”

As Varon noted, “The IRS has twice before failed to modernize.

And each time the IRS program fails, project management is seen



But the IRS story is not the only one getting attention from the technology press these days. Recently, a series of government and independent reviews have been undertaken to evaluate the FBI’s Trilogy program, a project launched in 1999 and slated for completion in October 2002, costing roughly $350 million. After 9/11, the FBI agreed to move the completion date to October 2002 at the prompting of the US Congress.

VOL. 5, NO. 7

4 By the fall of 2002, however, it became clear that the FBI and its contractors would not make the October 2002 date; moreover, it was likely that they wouldn’t make their original target of October 2003. The project was extended until the end of 2004, with funding increased by roughly $150 million. At that point, the FBI asked the National Academy of Sciences (NAS) to put together a team to review the project.3 Conducted by NAS, the evaluation dragged on for more than a year and a half, with its culmination in a recently published report suggesting that while the FBI has improved (it installed a new network and thousands of new desktop systems and hired a new government-savvy CIO), it still lacks a visible and viable enterprise architecture [2]. Meanwhile, the key new system in the works, the Virtual Case File, may not be ready for prime time. Even if it is, it should definitely be installed with a “flash cutover.”4 But like the IRS, the FBI has a spotty record in keeping its IT systems up to date. Perhaps more important, in the middle of the Trilogy project, the FBI was suddenly thrown into a frantic effort to support a major change in its mission from one of law enforcement to counterintelligence/ counterterrorism. While the FBI has been involved in domestic intelligence since its earliest days, 9/11 changed everything. Domestic counterterrorism

VOL. 5, NO. 7

AGILE PROJECT MANAGEMENT ADVISORY SERVICE became the number one priority. The NAS review team discovered that the FBI had continued with the Trilogy project without fully understanding what it was getting into. Although the FBI brought in top-level management talent from the National Security Agency, the challenge of developing stateof-the-art applications that integrate law enforcement and counterintelligence was far more difficult than anything previously attempted by the FBI. The difficulty arose from a fundamental conflict between the use and sharing of data between law enforcement and intelligence; while the first group wants to keep its data secret, the second needs to share. The NAS team pointed out that even after 50 years, this problem still plagues agencies like the CIA. THE USUAL SUSPECTS In other fields, whenever something fails, a detailed report follows that covers all aspects of the failure. For example, when the walkways at the Hyatt Regency Hotel in Kansas City, Missouri, USA, collapsed in the 1980s, an extensive analysis traced the failure to a design change that had not been reviewed properly. Summaries of these reports were made available to the press. When the Challenger and Columbia space shuttles failed, extensive reviews revealed problems with the design and decision-making processes. Once again, documentation of the

discoveries was made public so that engineers and managers could learn from previous mistakes. Out of these inquiries, the public learned about the dangers to these aircraft associated with O-rings and breakaway pieces of foam. But rarely do we find a very large IT project failure analyzed at this level, though one could argue that more economic damage results from IT failures than from building collapses or space shuttle disasters. Because of this lack of analysis, the little data we have on major IT failure always points to the same villains: project management, people, technology, or a bad contract. Usual Suspect 1: Project Management

Simply blaming bad project management is too easy to be of much use, largely because it involves a fundamentally circular argument. By definition, good project management means the project succeeds; and if a project fails, we have a case of “bad project management.” This kind of thinking doesn’t help much. Since there are many different kinds of project management, it is important to know which aspects of project management were not particularly good and which were lousy. Traditional waterfall methods, for example, use a series of steps. The idea is that if you follow the steps in the prescribed order, things will turn out fine. But



suppose we have a failed project and we did everything in the prescribed order. Is the cause bad project management or bad methodology? Agile development, on the other hand, says, keep things small, deliver early and often, stay on track. Not so many steps, but you follow them strictly and you get short-term feedback. The agile answer is not process but communication. But if a large agile project goes bad, what’s the problem? At what point does face-to-face communication break down? Usual Suspect 2: People

Next in the list of likely suspects in IT project failure is usually “pilot error.” The most common kind of airplane accident is referred to as CFIT, or Controlled Flight into Terrain.5 In other words, the pilot flew into the ground or a mountain or something else, thinking that everything was OK. In an aircraft crash, the first thing people look for is pilot error. In IT, we put it differently; we talk about project managers who are inexperienced or over their head or untrained in the latest technology. Often, the implication is that the project team didn’t understand what it was doing. Sometimes, the implication is that some of those on the project team were not working hard enough. But this is hardly ever the cause: no big project that I have observed, reviewed, or participated in has ever failed for


lack of effort. In failures as well as in successes, one of the hallmarks of VLPs is that you have dozens (or hundreds) of intelligent, earnest, hard-working individuals who work nights, weekends, and holidays to contribute to a goal to which they are committed. A mentor of mine at IBM once said, “No one at IBM ever gets fired for not working hard enough; they get fired for doing the wrong things.” People may be poorly trained or poorly motivated, but to paraphrase W. Edwards Deming, 85% of the time the problem lies in the system — that is, the approach — not the people. Usual Suspect 3: Technology

Consider these comments: “We just couldn’t get the XXX to work!”; “We would have made it if the technology had delivered the way they promised”; “We could never get the XXX to work with the YYY!” Technology often shows up on the review board’s radar when looking at significant failures. Fair enough. But again, my experience is that technology is rarely the ultimate cause of a VLP failure. Technology problems may aggravate an already bad situation, but technology is almost never the ultimate reason for things going wrong. However, betting on technology to bail you out of a bad schedule often creates serious problems.

Usual Suspect 4: A Bad Contract

And consider these comments: “We got screwed!”; “We didn’t pay attention to the contract”; “The vendor didn’t live up to the spirit of the contract”; “Given the contract, there was no way we could deliver what was needed.” Now we get to the ground rules: that is, the contract. Often times, the review board concludes that the project was doomed because it didn’t make the contract tough enough on the vendor. In selfdefense, the vendor usually responds that the customer didn’t know what it wanted, and by the time the customer figured out what it wanted, the vendor had gone past the time and costs specified in the contract. What gets lost in this discussion is the fact that contracts are twoway streets: if you block off traffic in one direction, things are liable to become congested. At this year’s Cutter Summit, one of the most interesting presentations concerned an approach that Cutter Consortium Senior Consultant Stuart Kliman called “Negotiation as if implementation mattered.” On VLPs, contracts are far more critical than on the average project because so much is unknown. Since VLPs come along so infrequently, few staff members have experience writing a contract that will span three or four years and that addresses multiple phases that involve defining exactly what the system will do.

VOL. 5, NO. 7

6 MORE THAN $I BILLION IN PROJECT FAILURES AND COUNTING So if the usual suspects are not the source of VLP failure, what is? If not project management, people, technology, and/or contracting, what should we look out for? What is the real problem? I’ve been pondering these questions for some time. I estimate that during my career I have been directly involved in reviewing and making recommendations on more than $1 billion in problem-ridden projects. When I add up the numbers, I’m shocked but not necessarily surprised. Nonetheless, I still find that I’m called in as an outside expert to “say grace” over a dying project. It’s a ticklish job, but you learn a lot. After a while, however, you can quickly tell how much trouble the project is in. Mostly, the smart people on the project have already deduced that the project is terminal; they just need someone to communicate this to management. But getting people to terminate an important activity is not easy or unique to IT. Recently, I found myself sympathizing with a doctor as he testified before a legislative panel in Florida and tried to explain why a young woman in a persistent vegetative state should be allowed to die. Try as he might, he could not convince a hostile group of legislators that just because the patient in question

VOL. 5, NO. 7

AGILE PROJECT MANAGEMENT ADVISORY SERVICE appeared to be aware and occasionally made gestures that seemed to represent smiles and frowns, she was nonetheless truly brain-dead and her chances for even partial recovery were infinitesimal. I know how the doctor felt. In the vast percentage of what I now call “permanent vegetative IT systems,” I end up recommending that the project in question expire. In the next section, I offer lessons learned about killing large IT projects. In the final sections, I pull together what is unique about VLPs to help you avoid making the mistakes that have led to costly disasters. PROJECT TITANIC SCENARIO All big failures look alike. — Ken Orr

Looking back, the most surprising thing about all the failed projects I have studied is how much they resemble one another. When I researched the history (via logs) of some of the failed projects I worked on early in my career, I found that they look enormously alike — so much so that my first book on systems development describes the following 21-step approach for large project failure [6]. I refer to this multistep approach to project failure as the Project Titanic Scenario:

Step 2: Select acronym (i.e., set project scope). Step 3: Select project manager. Step 4: Develop project schedule. Step 5: Hire requirements analysts. Step 6: Develop requirements specifications. Step 7: Develop input specifications and order data input forms. Step 8: Hire programmers. Step 9: Begin work on input edit programs. Step 10: Begin work on conversion programs. Step 11: Begin design of database. Step 12: Begin work on update programs. Step 13: Hire additional programmers. Step 14: Begin work on output reports, query screens. Step 15: Slip schedule. Step 16: Hire additional programmers. Step 17: Panic. Step 18: Search for guilty.

Step 1: Select project due date and budget.

Step 19: Punish the innocent.


EXECUTIVE REPORT Step 20: Promote the uninvolved. Step 21: Go back to Step 1. It is depressing that more than 25 years later, I can still present this scenario to a large group of IT managers and professionals and someone will invariably approach me and remark, “I think my project is at Step 15” or “How do we get off this project before we reach Step 18?” Most large failed IT projects look alike, and the causes of failure are eerily similar. (Note: To those who have dismissed the Project Titanic Scenario as “cute” or “trite”; it is neither. Built into the scenario are the seeds of the failure of VLPs.) Recently, I have received several RFPs that could have been drafted directly by following the Project Titanic timeline. So in the next few pages, I will try to explain why what many (i.e., most) people would consider “straightforward project management” fails when applied to VLPs. I do so by relaying observations drawn from personal experience. I have already touched on some of them in our discussion of the usual suspects, but they are important to reiterate. Observation 1: Large projects don’t go bad; they start bad. The root cause is not technology, but ego. On most failed projects, two egos create the conditions for failure: the ego of the sponsor (aka the boss) and the ego of the project manager (aka the fall guy).


For some reason, top managers and executives believe that they can do things with IT projects that they would never attempt in other kinds of projects (e.g., building a plant or rolling out a new product). In itself, this is not an earthshaking observation. Executives are always in a hurry and always want to know the costs up front. And executives, especially top executives, all have big egos. Unfortunately, on large IT projects, several factors often conspire to drive project risk through the ceiling and a large percentage to fail. The most important of these factors is the ego of the project manager. On one failed project after another, we find project managers with personality traits that sow the seeds of project failure. On a large project, the typical project manager is ambitious, articulate, proud, thrill-seeking, inexperienced, and a team player. In other circumstances, including smaller IT projects, these personality traits yield success; but on large projects, they tend to spell disaster. In my experience, it takes two to create a really big failure. It is not just the sponsoring executive’s ego and desires that set up a project for disaster, but that these demands are coupled with the ego of the project manager, who has a do-or-die attitude. Note that in the Project Titanic Scenario, the project manager isn’t even hired

until after the most critical decisions (i.e., the project deadline, budget, and scope) have been made. This means that the project manager is willing to accept the mission impossible and make it happen. Unfortunately, real life and $100-million movies often turn out differently. Observation 2: Most of the critical bad decisions on large, failed IT projects are made early on. Many people, especially project managers assigned to failed IT projects, believe that if you’re smart enough and work hard enough, you can overcome any obstacle. Unfortunately, this is patently false. A project’s fate is often determined in the first few weeks of the project’s conception, often before it is even officially a “project.” Mary Shaw, a friend of mine at Carnegie Mellon University, once recalled the ideas of one her colleagues, a chemical engineer, on the critical nature of these early decisions. He said, “If you were building a new chemical plant, the expenditures tended to be ‘back end–loaded’ — that is, most of the costs were associated with the construction of the plant.” (See Figure 1.) This same cost curve shows up in a large number of development projects as well. “But,” continued Mary’s colleague, “if you look at when these expenditures are committed, the graph is very different. Almost all

VOL. 5, NO. 7



the expenditures are ‘committed’ [i.e., decisions have been made] very early on in the project, in the first weeks or months.” (See Figure 2.) Here too, there are strong similarities with IT projects. Moreover, the chemical engineer’s comments about cost also apply to all other aspects of a project, including project staffing, project scope, methodology, tools, and schedule.

Successful project managers know this; unsuccessful ones don’t. On projects that follow the Project Titanic Scenario, the budget, end date, and scope are, for all intents and purposes, established before the project manager is even selected. As mentioned previously, this is a recipe for disaster. Unfortunately, the knowledge of the application and/or



Time Figure 1 — Project expenditures over time.

technologies and methods that will apply to the project often lags significantly behind the decision making. The middle curve in Figure 3 is intended to demonstrate that for most projects, knowledge acquisition is “lumpy.” Normally, a great deal is learned about the application domain during the early part of the project, and another spurt of knowledge occurs late on the project. But for the most part, this knowledge is not factored into the major commitments. Experienced project managers understand Figure 3 intuitively. They know that they have to be very careful about committing too much too early. They know that the more you know, the easier decision making becomes. As I will discuss later, architecture is crucial, because on large projects, a great deal of overhead is associated with knowledge management (acquiring, documenting, and disseminating knowledge about the project design). THE WRONG PEOPLE

Expenditures committed



Anyone who’s been on one large failure will make sure that they are never on another. Therefore, most large projects are staffed with trainees and masochists. — Andy Kinslow, IBM

Time Figure 2 — Commitment of project expenditures.

VOL. 5, NO. 7

In most areas of human endeavor, if you’re going to take on a project that is likely to cost tens or hundreds of millions of dollars and



One of the excuses frequently offered for a lack of project leader experience is that older project managers don’t understand the new technologies and/or methods. This may be true; old dogs have trouble learning new tricks. But if this appears to be a major issue, it should serve as trigger: do you really want to set out on a critical undertaking using technologies and methods that only a few young people really understand?6 Observation 3: Failed projects are often managed by those who are inexperienced in managing VLPs with the new immature technologies and methods. I have found that the key to the success of any project, especially a large project, is having the right people with the right skills in the right jobs. I wrote a series of Cutter Consortium E-Mail Advisors on the five big questions that every organization should ask


Expenditures committed Domain/technology knowledge Knowledge


that will have an enormous effect on an organization, you must insist that the project leader has extensive experience working on a similar project. Moreover, you will check credentials, and closely. You expect the project leader to have a professional license and to have worked in the field for an extensive period. Because the ramp-up for VLPs is often rushed because of a fixed deadline, people frequently forgo this process when it comes to large IT projects, which leads to major problems.

Time Figure 3 — Expenditures, commitments, and knowledge.

about those selected to lead large, critical projects [3, 4, 5]. These questions address experience and risk: 1. Has any project team member built anything this big? 2. Has any project team member used this methodology/ technology on a project this big? 3. Has any project team member been involved in a large systems failure? 4. Do the right people have planning, architecture, requirements, and design skills? 5. Has anyone thought about contingency plans? In order to know whether a project team has enough experience to warrant starting a big project, you need to ask probing questions. With all the major positions represented, this questioning should be more than cursory. It

should address details and ask for a description from the candidate of what he or she does when things go wrong or priorities change. This questioning should check how the applicant might choose to produce short-term deliverables. The candidate should be grilled on his or her experience with large projects. How large a project has the candidate run or worked on? Big projects require specific skills in planning, scheduling, delegating, and monitoring. Successful project managers know that there is enormous pressure to do dumb things, and it may be hard to retain top management’s attention, especially if the project is going to last for a few years. And it is not enough for IT project managers just to understand the intricacies of project management. Indeed, most project management training doesn’t prepare project managers for a VLP with

VOL. 5, NO. 7

10 its unique problems. Indeed, I am not a fan of “content-free” project management training; I’m not sure what it means to be a “certified” project manager if you haven’t actually managed a large project yourself. Managers of VLPs must know what their people are doing and why. When managers can’t understand what their people are saying, they are at the mercy of their people. Large project failures often involve one of two scenarios: (1) the project manager fancies him or herself a “pure” project manager and puts all trust in the chief designer, architect, or perhaps a consultant; or (2) the project manager is an advocate of some new approach or tool. Both scenarios present difficulties. Observation 4: On large projects, there are no silver bullets. While advanced technologies and methodologies are often treated as silver bullets that can magically overcome one or more major project management problems (e.g., time, cost, and personnel), relying too heavily on them is always an indication that something is wrong with the project planning. Over the years, I have watched managers on one failed project after another put up a smokescreen by saying that they were implementing new methods and tools to avoid embarrassing questions about schedule and process.7

VOL. 5, NO. 7

AGILE PROJECT MANAGEMENT ADVISORY SERVICE Observation 5: On a VLP, there is no substitute for a project manager who has a failure under his or her belt. You’ll notice that the third big question is about the importance of failure. If the project manager has not been involved in at least one major failure, management ought to be cautious about launching the project. I tend not to trust people who have only experienced success, because, to paraphrase Henry Petroski, one of the great writers on engineering, you simply don’t learn enough from success [7]. Failures provide a template in the mind of a project manager. If you’ve survived a major failure, when you hear certain phrases or see things that happened on the failed project, you instantly become more alert and start asking more questions. Observation 6: VLPs require experienced project architects and project managers. Over the years, I have tried to be both the project manager and the project architect. It’s a tough job, and the larger the project, the more difficult it becomes. Either one can be the ultimate boss, depending upon the kind of project, but in general, you need two different people: one to worry primarily about the system that you’re trying to develop and the technology being used to install it (the project architect) and the other to worry primarily about communicating with management, producing the right reports, and tracking the

team’s progress (the project manager). Finding a good project architect is every bit as difficult as finding a good project manager. In fact, good project architects are probably more critical to VLPs than are good project managers. They must have vision and understand the details. They need to know when programmers are lying or fudging. They need to know what people should be testing. They should constantly think about integration. And perhaps most important, they need to know when to blow the whistle. Project architects are the ones who worry about the project’s scope. THE WRONG APPROACH On a VLP, other than putting the wrong people in charge, the worst thing you can do is use the wrong approach (methodology). If you look at the Project Titanic Scenario carefully, you will notice that it is actually a distillation of a series of the right tasks, but in the wrong sequence. Project Titanic involves setting the scope too early, hiring too many people too soon, and starting them working in parallel on the wrong end of the problem (i.e., the inputs) when you have no idea what the outcomes (i.e., the outputs) should be. And it involves breaking up the project in phases only when you’re in deep trouble. But fundamentally, Project Titanic is just a natural reaction to doing



the very worst things first — namely, establishing the due date and the budget too early. Most inexperienced project managers believe that if your due date and budget are set ahead of time (i.e., before the project begins) the best you can do is develop a schedule working backward from the due date, breaking up the tasks and sequences as much as possible so that much of the work can be done in parallel, and then hoping for the best. As it turns out, this approach is nearly always a fatal mistake. Starting from the given due date is not the best you can do; it is actually the worst. It just allows you to make management happy initially and postpone the most serious problems, at least for a while. But sooner or later, the project will start slipping and you will ultimately have to go back and ask for more money, more people, and more time. And eventually, if you keep going back and everything looks hopeless, they’ll call in someone like me. At that point, the task is to say what project leaders failed to say at the beginning — the truth. Observation 7: The initial project scope of a VLP is never right. You simply can’t set the project scope correctly until you know the basic requirements and general design, so don’t try to set scope in concrete too early. A great many people believe that “hard-headed”


project management consists of deciding on project scope early and then religiously fighting to avoid scope creep. The notion is a good one, but it doesn’t work. Recall Figure 3: people on failed projects tend to make their most important (and most fatal) decisions before they have the knowledge to make those decisions. This is especially true of project scope.8 Take the following example. Suppose our scope is “to install a new supply chain management system.” Lots of million-dollar systems are launched with that brief a goal. But what if that’s not our problem? What if the reason we have such difficulty with our supply chain is that we have a such terrible accounts payable system (or process) that people don’t want to do business with us? Assuming you can set the right project scope without doing requirements and high-level design is like assuming you can plot the route of a new highway without surveying the land. Sure you can draw a line on a map and say, “I want a road from here to here in three years.” But suppose that when you performed the land survey for this new road, you found that the proposed route crossed a river, a swamp, and a couple of mountains. What would you do? The answer is you would change your scope (or time frame), and quickly.

Inexperienced, content-free project managers (i.e., those who believe that all you need to know about managing a large software project is project management itself) think that being hardheaded means that you make all the important decisions, then build a plan and get users to “sign in blood.” But often a wildly optimistic plan will rarely bring in an unspecified project before the arbitrary target date and under the arbitrary budget. This may not a big deal on a small or medium-sized project, but on a big project, it’s a killer. Observation 8: Check your maps frequently. All large failed projects have Gantt/PERT charts, and none are up to date when things go south. The map is not the territory. — Alfred Korzybski If you don’t know where you are, a map won’t help. — Anonymous A Gantt chart is what you show management while you’re doing something else. — Ken Orr

Every large failed project that I’ve ever reviewed has had an elaborate project Gantt/PERT chart early in its history. So when I start gathering data, one of my first questions is, “Where are we on this chart?” And by this point in the project, the response is always,

VOL. 5, NO. 7



“We can’t actually show you where we are on this chart” or, even worse, “Which chart?” During the first six or 10 months of a big project, the project charts are religiously updated and status reports are filed right on time. But as the project begins to fall behind, the scheduled updates become more sporadic and the charts less reflective of “the plan.” One of the reasons is that catchup strategies are often introduced, and it becomes too involved, not to mention embarrassing, to actually update the schedules and dependencies to reflect the revised situation. Observation 9: Project managers and their bosses will go to great lengths to avoid telling top management the truth. The truth will make you free, but first it will make you miserable. — Tom DeMarco

In most failing projects, there are a few people at the top of the organization who think they are in trouble, lots of people at the bottom who know they are in trouble, and a bunch of worried middle managers trying to keep those at the top from talking to those at the bottom. The hardest thing to come by on a big, problem-riddled project is the truth. This is understandable. In all these VLP cases, dozens, even hundreds, of people have worked night and day on the project for

VOL. 5, NO. 7

years, so admitting that a project is in serious trouble is akin to admitting personal failure. But in the end, truth is the only guide. THE END GAME: KNOWING WHERE YOU REALLY ARE In 1707, four English Ships of the Line went aground on the Scilly Isles off England’s Southwest Coast. Their captain thought that they were two hundred miles further East. Two thousand sailors were lost in this tragedy. — Dava Sobel [9]

Observation 10: Project schedules on big projects do not slip one day at a time; they slip in big chunks. In one instance, I was involved on a very large, highly critical, extremely overdue project. The situation got so bad that management began to hold daily progress meetings. Every afternoon, management convened all the key managers and reviewed project progress in detail. One day the news was particularly bad. “How the hell did we slip six months overnight?” the senior VP screamed. “We worked very hard, Sir,” replied a small voice from the back of the room. Of course the project didn’t slip six months overnight. Like Sobel’s captain of the English fleet, the project team members didn’t know where they were or, more to the point, how much still had to be done. What they discovered was that they were much further from their target than they

thought. They had been operating on the assumption that they were located at one place on the map, while they were at another. Or more embarrassing, they realized they had the wrong map! If you haven’t developed requirements and general design, you can’t estimate what you have yet to do in detail. When you finally determine how much work remains, the amount often totally exceeds what management had been promised, even in the most recently revised schedules. Observation 11: In software development, you can lie about everything but the testing. Testing for VLPs is like the fatal shoals on which ships run aground. Project plans, requirements, design documents, program charts, and so on are just pieces of paper. Unless they are reviewed carefully by those who know what they’re doing, they are not really deliverables. If they choose to, project teams can hide major problems under mounds of paper. But ultimately, you can’t lie about testing. If a program doesn’t work, it shows up in testing — you can’t go on; and if the system does work, testing is the final arbiter. Many large projects maintain the lie that they are on schedule right until they start serious testing. Unfortunately, on large projects, many of the most critical tests — the ones the end user sees — don’t occur until it is way too


EXECUTIVE REPORT late to change course.9 The result is like hitting a large rock. The project careens off schedule and stays off schedule. Integration schedules slip and data conversion can’t progress. Meetings are held daily, and one excuse after another is offered, the search for the guilty commences, followed by the punishment of the innocent, and so on. Observation 12: Most VLPs are not cancelled; they simply slip away. Admitting that you blew (or were party to blowing) millions of dollars and wasting years of time is never a pleasant task. As a result, all but the most visible large projects are labeled “partial successes” and allowed to disappear from management’s radar. Whenever the project comes up, there is a wink and a nod and the subject changes. Years or even decades later, people change the subject when the project is mentioned. To keep the discussion muted, those most closely associated with it are allowed to retire or slip into staff positions. They’re not fired; they just fade away. Occasionally, things get so bad that the company is hemorrhaging money and those involved become so obstinate that someone like myself gets called in to put the project out of its misery. And even then, the project is not declared a failure. No wonder we have so little data on these kinds of projects.


WHAT ARE THE REAL PROBLEMS WITH VLPs? I have often asked myself if anyone really knows how to manage VLPs. Lots of people, and organizations, claim to know how and that they routinely bring big projects in on schedule and under budget, but I’m skeptical. Given my experience with VLP failures, I am often less than impressed when I hear about project successes, largely because it is difficult to know just how successful any VLP was unless it was reviewed after the dust settled, preferably in a post-audit review a year or two after the declaration of project success.10 People Don’t Want to Remember

So the success or failure of big projects is difficult to judge up close. Moreover, the trauma of being involved tends to cloud our vision. Here’s a personal example: Many years ago, my wife and I remodeled our 19th-century Victorian house, which was a major undertaking. We added a family room, an office, a new master bath, and redid the kitchen. The renovation involved almost every very expensive thing you could do to an old house. The remodeling took months, and we moved out of the house (something we hadn’t done the last time we remodeled and learned to regret). And though we worked through every possible detail (or so we thought) with the

architect over the previous 18 months, the job cost far more than the estimate, and naturally everything was late. And every time we asked the contractor when a particularly important milestone such as the installation of the cabinets would be done, his response was “two weeks.” But somehow, the contractor’s two weeks had a habit of dragging on for months, and things got more and more expensive. Shortly after the remodeling was over, my wife and I needed some relaxation, and we went to see the comedy The Money Pit. In the movie, a couple buys a run-down mansion and remodels it. Like us, everything the couple touched fell apart, and — again, like us — when the couple hired contractors, everything was scheduled to take “two weeks.” The costs, like ours, shot up. It was a funny movie, I suppose; everyone in the theater laughed hilariously. But somehow my wife and I did not find the situation funny. The Money Pit was too close to the truth for us after going through the real thing. That’s the way it is with traumatic events; people block the unpleasant experiences from their mind. And so it goes with people recovering from the shock of a very large failure. Often, when I mention the Project Titanic Scenario, I see tears come into people’s eyes. The rule is, if you were involved on a recent VLP, Project

VOL. 5, NO. 7

14 Titanic isn’t likely to be funny. And unlike The Money Pit, there is rarely a happy ending. Seat-of-the-Pants Management Doesn’t Work

VLPs fail so frequently because many of our most basic (instinctive) approaches for managing people, teams, and organizations that work (or appear to work) on small or medium-sized projects lead us astray as a project exceeds a certain size. In most organizations, most people manage most of the time by the seat of their pants. On VLPs, however, this kind of management is a bad thing. To demonstrate, let’s trace the origin of the expression “by the seat of one’s pants.” Early airplane pilots learned that people have an instinctive sense of the directions of up and down. In most earthbound circumstances, this sense of up and down is both correct and useful. In flying, however, our natural senses work against us. A large number of pioneering pilots were killed before it was discovered that when you’re flying an airplane, you can’t trust your instincts. We now know that a combination of centrifugal forces and a lack of visual cues often trick inexperienced pilots into believing that up is down and down is up, resulting in the previously mentioned CFIT experience. 11

VOL. 5, NO. 7

AGILE PROJECT MANAGEMENT ADVISORY SERVICE But it turns out that it is hard for your brain to override these instincts. And for beginner pilots, the hardest task is learning to trust the instruments in all cases. As pilots will tell you, the most demanding part of learning to fly is becoming “instrument rated.” During this training, the pilot is blocked from seeing out of the airplane and must give up the reliance on instinct concerning direction and learn to trust what the instruments tell him. Because the skill is so difficult to master, it is not necessary to become instrument rated to get a basic license to pilot a plane. But a pilot who isn’t instrument rated is not supposed to fly when visual flight rules12 do not apply. And because of the real dangers involved, most noninstrumentrated pilots are careful to avoid flying in conditions when they lack visual clues. But occasionally, changes in weather, time of day, or location conspire to place the unrated pilot at risk or — as was the case with John F. Kennedy Jr. — tragedy. In the same way that our physical instincts betray us when flying an airplane without visual clues, our management instincts have a way of betraying us when trying to manage a project for which we have to trust our “project instruments” rather than our instincts. Small or medium-sized projects can be managed by direct communication and/or inspection. With small or medium-sized

projects, there are only a couple of levels of management at most. But with large projects, and especially VLPs, there are layers upon layers of management to the point that it is impossible to use the normal sets of visual clues that indicate if the project is flying up or down or where you are with respect to the project terrain. (In project management, as in flying, it is important to know where you are with respect to things like airports, coastlines, the surface of the earth, and mountains.) Not only does managing by instinct cause large projects to veer off course, but many of our basic planning assumptions often fail us on VLPs. The most important of these assumption is what I call the “front-to-back strategy.” Designing Front to Back Doesn’t Work

In general, there are three basic strategies for developing large systems: 1. Front-to-back strategy. Start with the inputs and work “forward” to the database, then to the outputs. 2. Back-to-front strategy. Start with the outputs and work “backward” to the database, then to the inputs. 3. Middle-out strategy. Start with the database (or central objects) and work backward to the inputs and forward to the outputs.



Of these three approaches, the front-to-back strategy is by far the most instinctive and intuitive. Humans tend to think linearly and temporally, and we take our clues from execution. When this system is done, so the thought process goes, it will start with the inputs, update the database, and then periodically produce outputs, so that’s the way we ought to build it. Unfortunately, the front-to-back approach, like seat-of-the-pants management, is a trap that eventually catches up with you, usually late in the project when you can least afford to redo things. Here’s the problem.

find a place on the input to place that information. On VLPs, however, the cycle of designing the inputs, the database, and the outputs is very long, and if you discover, as you always do, that someone missed one or more of the major outputs in requirements, you need to make major changes to both the database and the inputs. This “refactoring” can throw a monkey wrench into the planning process.

The reason we do systems is to be able to produce certain output (payroll checks, up-to-date bank statements, benefit checks, etc.). Those outputs comprise data elements that are produced either from inputs or from data fields computed from inputs, together with various calculation formulas or decision rules.

Working front to back, however, looks good from a project planning standpoint. IF we are constrained (from a project management standpoint) by an arbitrary due date, and IF we can define the inputs early, THEN we can define the database early. And IF we can define the database early, THEN we can have lots of people entering data, AND we can also get lots of people writing output reports independently of one another. I refer to this as “fantasy project planning.”

In small or medium-sized systems, the connections between inputs and outputs are usually pretty clear and direct. That means if you start with the inputs, put those inputs on a database, and produce outputs from the database, the process doesn’t take long. If you make a mistake and find that you can’t produce a given output from the data you have on the database, you usually have time to go back and add a new element to the database and then

But of course, the real world does not follow this pattern. We almost never know what all the primary outputs for a large system are at the beginning of a project. Often, the reason we’re tackling a big system is that we need to change things. That means defining new requirements, and new requirements normally involve new outputs or new business rules or both. So we put aside the most important consideration — namely, defining the primary


outputs — and start guessing about what inputs we will need. Once we come to a compromise on the inputs,13 we guess what the database should look like and then finally define the outputs. Then we discover that we need things to produce the outputs that no one told us about, so we have to go back and refactor. And all this at the worst possible time. Learning to take a back-to-front approach, on the other hand, is much like learning to trust your flight instruments. It is hard to concentrate on defining the outputs when there is pressure to get started implementing the system, but a back-to-front strategy works much better on large systems. The great advantage of a noninstinctive, back-to-front approach is less rework, and in many big projects, rework is the killer. If you do a good job of defining the primary outputs of the system, designing the database becomes much more straightforward, and defining the inputs is a piece of cake. As a result, the implementation goes much more smoothly. But unfortunately, using a backto-front strategy requires patience, and patience is often in short supply on VLPs. Starting Fast Doesn’t Work

In the absence of long-term plans, short-term demands will always drive out longterm goals. — Ken Orr

VOL. 5, NO. 7

16 In the Project Titanic Scenario, the due date and budget are already assigned when the project manager is selected. Surprisingly this is often the case in the real world as well. As a result, many, if not most, project managers are already late when then they get their assignments. Therefore, there is great pressure to get cracking and to cut corners. And when, in our boss’s perception, we’re already behind, our instincts tell us to start fast and rush. But here again, our instincts get us in trouble. On VLPs, the tendency, especially among inexperienced project managers, is to confuse effort with progress, which is natural. Human beings tend to be impatient. In this age of instant gratification, impatience is compounded with the illusion of quick fixes. Big projects require careful planning, but on many large projects, there is a not-so-subtle message that there isn’t any time for planning, only execution. On small or medium-sized projects, this can be an annoyance, but on VLPs, confusing effort with progress is deadly. Based on the fast-start scenario, there is often a push to bring people on too early, and once they are on board, there is ongoing pressure to keep everyone busy. The result is that the project manager often struggles to stay one day ahead of the programmers. At some point, something falls between the cracks and everything begins to slip.

VOL. 5, NO. 7

AGILE PROJECT MANAGEMENT ADVISORY SERVICE The truth is that long-term things like planning and architecture are the keys to success on VLPs, and typically, these steps can’t be accomplished by large groups working in parallel. The early parts of VLPs must be serial activities, and they must be done well before you bring people on board. If you are a manager of a large project, people will test you. Lots of top managers believe that if they set totally unrealistic schedules and demands, people will perform better. But the opposite is more often the case. Most people work best when they believe management knows what it’s doing and can set reasonable expectations. Because I’ve seen the premature start strategy fail so often, I get really testy when people attempt to bully me into “doing something even if it’s wrong.” WHAT SHOULD YOU DO? If you’ve read this far, you know that there are many ways that VLPs go south. What I’ve done is describe the landscape, and I’m not going to turn around now and say that a cookie-cutter approach will work, because it won’t. The problem is that we have too many constraints. If you’re supposed to deliver an enterprise-wide supply chain system for less than $10 million by 1 September 2005, your project may already be doomed. Moreover, it doesn’t matter much whether you use a traditional waterfall methodology or an agile

approach; the cards are stacked against you. This section discusses various activities that can dramatically alter the odds in your favor. And if you are committed to using agile development methods, it gives you a framework. Break the Project in Two

In the real world, there are architects and there are general contractors, and they work together in different ways, however, logically the architect always comes first. The architect does all the basic design work, checks out the lay of the land (literally), reviews rules and regulations, and meets regularly with the owner and/or developer to establish the specifications. The architect produces the building specifications in terms of a variety of sketches, models,14 and drawings. Historically, the architect receives 10% of the cost of the entire building, so if the cost of the building is $5 million, the architect gets $500,000; and if the building is estimated to cost $50 million, the architect gets $5 million. Architecture is recognized as a major component of the entire cost structure. Moreover, it involves detail as well as highlevel design. Most important, from our viewpoint, the funding for the project is usually broken into two pieces: (1) the planning and architecture and (2) the construction. If we are to remove the risk from the management, we must get our



knowledge in sync with the magnitude of our decisions. Phrasing this new approach in terms of risk management is often easier for executives to grasp. They may not understand VLP management or knowledge management, but they certainly understand risk management. Do Requirements and Architecture First

The division between architecture and construction has worked for hundreds of years in construction, and I believe that we must employ a similar strategy on VLPs in the software world. If we are ever to feel comfortable managing VLPs, we must focus on understanding the overall design (high-level requirements and high-level architecture) before we break the project into its major components and develop detailed schedules and detailed cost breakdowns. So we must separate the high-level requirements and architecture from the development (i.e., construction). This is just as important in agile development as it is in a traditional waterfall environment, perhaps even more so. To work well, agile development needs to break a large project into a set of small projects (small enough to be done by small teams composed of dedicated users and two programmers). In the next section, I will describe how this might be accomplished in the context of a VLP.


Make the Architect Responsible for Quality Control and Integration

In the real world, architects do more than design the building; they help estimate the cost and schedule, and equally important, they are responsible for change control and quality control throughout the project. I think they should play this role in VLPs as well. And if you break a large project into several small (i.e., incremental) subprojects, you must ensure that, in the end, they fit together to create the finished product. On a VLP, a project architect should be responsible for the cohesion of these pieces as well as for change control and quality control. Since computer science has not reached the level of professionalism of architecture and other engineering disciplines, it’s not likely that we will be able to pull off really big projects routinely for some time, but these kinds of changes could make a substantial dent in our VLP failure rate. SHUTTING DOWN AND STARTING OVER Things to Do on a Failing Project

If you are ever asked to go to another organization, or another division of your organization, and evaluate a large project that is in trouble, you can use some of the observations presented here as a sanity check. You should know that, in most cases, by the time an outsider gets called in, it’s

too late to do anything but conduct as quiet a burial as circumstances allow. But the process still requires careful analysis, courage, and tact. And no matter what, submit your report to top management so that it is a permanent part of the project record. Start with the people on the front line: the designers, programmers, testers, and documenters. Have management construct a current Gantt chart. Have an all-hands meeting and distribute sealed ballots. Ask everybody to write on the ballot the probability of the project meeting its goals. Even under the worst of circumstances, most people will be honest, especially if they believe their anonymity is protected. Look under rocks and look the project managers in the eye. Don’t be afraid to recommend that the project be scuttled. Remember, everything that has been spent to date is sunk cost. Economically, it doesn’t matter how much has been spent; what matters is how much will likely be spent going forward. Most organizations don’t like to advertise failure; it’s bad for the career, you know. But keep all your notes from the project so the next time you’re asked to be a project manager on a large project, you can pull out your notes and check them against the project schedule and due date. If the situation resembles the failed project you just helped kill, start

VOL. 5, NO. 7

18 making a fuss early. If nobody listens, start updating your résumé. Recall Andy Kinslow’s comment: large projects are often manned by trainees and masochists. Which one are you? Starting Over

Let’s suppose that you have successfully shut down a VLP that failed. What next? Well, if the organization is game for another round, it may be willing to consider new alternatives. (And if management wants to set another due date and budget, run — don’t walk — away!) On the other hand, if the organization is willing to rethink its basic strategy (and organizations that have recently admitted failure are often more humble), suggest the concept of the project architect and breaking up the VLP into two major parts: (1) architecture and (2) construction. If managers think these ideas make sense, explain that they need to think of committing to the architecture first and then reassess how much the construction phase will cost and how it will be staged. If they are still listening, tell them about agile development. Leveraging the Strengths of Agile Development Combined with Project Architecture

Agile development doesn’t have much to say about overcoming the special problems of VLPs, because few VLPs have been successful with agile. While agile VOL. 5, NO. 7

AGILE PROJECT MANAGEMENT ADVISORY SERVICE development gurus refer to fairly large projects, they’re not usually talking about VLPs in our sense of the word. Agile development performs well in small and medium-sized projects. Based on the approaches and methods of Cutter Consortium Senior Consultants Kent Beck, Alistair Cockburn, and others, the industry is rethinking how projects should be conducted. The emergence of agile development has prompted organizations to reconsider how people communicate on real projects and caused many to recognize the need to eliminate the busywork and to get all the major parties in the same room working on the same end product. On VLPs, however, it’s difficult to get all the right people in the same room. And even if you did, it’s unlikely that much face-to-face communication would take place. If agile development is going to scale up, it clearly must come up with some major enhancements. I strongly believe that ultimately a lack of architecture will plague large-scale agile development projects just as it has traditional development projects. Indeed, in discussions on the use of agile development on large projects, many individuals have indicated to me that they insert an architecture phase routinely as the first step. We all know that large projects increase both the opportunities

and the need for communication faster than they increase knowledge. The larger the project, the greater the number of people who need to know about the various aspects of the design, which elongates the communication cycle, which creates the need for more meetings, and so on. Inserting an architecture phase attempts to maintain the balance between knowledge and effort. Figure 4 offers an example of a relatively large agile development project developed using the Feature Driven Development methodology in Australia. The project was broken up into several subprojects, each of which can be accomplished with agile. The secret here was that before work began on the subprojects, a fairly extensive architecture phase broke the entire project into smaller ones. But breaking things into manageable pieces is not enough for an organization to be successful on really big projects. Agile development solves three major issues: (1) how to break a big thing into a set of smaller interrelated pieces; (2) how to integrate the separate pieces back together; and (3) how to accommodate changes that affect the interrelation between the pieces. The second and third issues pose significant difficulties. Adding an architecture phase to the agile lifecycle solves only the first of these problems. In order to



solve the other two, more has to be done; we must deal with the issues of change/quality control and integration. What needs to be put in place, then, is a structure somewhat like the one in Figure 5. Architects in the software/systems world ultimately will have to assume responsibility for their architecture all the way through implementation just as they do in real-world construction. Eventually, one might guess that project architects will be licensed, just as real-world architects and engineers are today. At any rate, for project architects to help solve our problems, they must closely monitor the development to ensure that what was originally designed gets built and, if material changes are made, what the consequences will be. Tools, Communication, and Online Collaboration

Subproject subproject11

Subproject subproject22

Subproject subproject33

Architecture Architecture

Subproject subproject44

Integration Integration

Subproject subproject55

Subproject subprojectnn

Figure 4 — Agile development framework for a very large project (VLP).

Subproject subproject 1

Subproject subproject 22

Subproject subproject 3

Planning Planning

Architecture Architecture

Subproject subproject 4

Integration Integration

Installation Installation

Subproject subproject 5

In the end, traditional waterfall development schemes, such as the CMM®, and agile development have much to learn from each other. CMM must learn to be more focused on the product and less on the process, while agile must learn to be more focused on architecture and on a formal means of documentation. I believe that, in time, much of this argument will be resolved through better tools for requirements, architecture, design, and collaboration. In the end, distance will be less of a barrier to communication.


Subproject subproject nn

Integration and change management

Figure 5 — Augmented agile development framework for a VLP.

Since my company, the Ken Orr Institute, is a small consulting organization with widely dispersed clients and coworkers, the collaboration tools available on the Internet allow me and my

coworkers to communicate with a wide variety of others (clients, researchers, vendors, etc.) in ways that would have required substantial travel and/or wait time in the past. I believe strongly

VOL. 5, NO. 7

20 that CADD/CAE/CAM tools for IT will permit all organizations to attack VLPs in an increasingly agile manner. I have something of an obsession with tools not because I want to replace people but because I fear that unless we find better tools for developing, maintaining, and operating large systems, our organizations will become more and more difficult to manage. Already, many organizations are so dependent on a few systems and networks that are fragile, possibly too vulnerable to survive major business or technological changes. CONCLUSION No one knows how to bring in VLPs successfully every time, even under the best of circumstances. What I do know, based on thousands of hours and dozens of projects, is what not to do. Managing VLPs takes great management and architectural skills. Until recently, people have underestimated the importance of architecture in VLPs, and we must redress that. Before we make the critical decisions in these projects, we need knowledge, and the best way to get high-level project knowledge is through architecture. Remember: hope is not a strategy. EPILOGUE Life has a way of repeating itself. Many years ago, I got an interesting phone call from someone at the IRS. After reassuring me that

VOL. 5, NO. 7

AGILE PROJECT MANAGEMENT ADVISORY SERVICE the call didn’t concern an audit, the caller said that a colleague of mine who worked in the IRS’s methodology section had suggested that he give me a call. “We have a big problem,” he said.

Probably 15 years have passed since this conversation, but clearly the IRS has still not been able to modernize its core system. During this period, the IRS has gone through at least two, and possibly three, iterations at roughly $8 billion a whack.

“How big?” I asked. “Eight billion dollars in 10 years,” he said. “More like $20 billion in 20 years,” I calculated silently to myself. Out loud I said, “That’s big.” “What kind of methodology works in a project this size?” he asked. “None known to man,” I said. “What should we do?” he said. “Well, I’ll tell you, because I know you won’t do what I suggest. What I would do is hire 10 or 12 of the brightest people you can find inside and outside the IRS and put them in a room for about 12 to 18 months. I’d have them come up with a plan. Then I would start spending that $8 billion. By the way, my guess is that to get the job done, it’s going to take more like 12 to 15 years.” My comments met silence on the other end, and then the IRS worker said, “You’re right, we’re not going to do that.” We then discussed the ideas the IRS had for spending the $8 billion for a few more minutes, and then the IRS worker rang off.

The phone call, brief as it was, got me thinking about architecture and VLPs. It has occurred to me more than once that the time, cost, and organizational pressures of VLPs are such that it is nearly impossible to succeed. Somehow, the cycle must be broken. Here I am, 15-plus years later, talking about what the IRS should do. I think I was right then and that I’m right now, but I’ll give you even money that the IRS is still not likely to follow my advice. ABOUT THE AUTHOR Ken Orr is a Fellow of the Cutter Business Technology Council and a Senior Consultant with Cutter Consortium’s BusinessIT Strategies and Enterprise Architecture Practices. He is also a regular speaker at Cutter Summits and symposia. Mr. Orr is a Principal Researcher with the Ken Orr Institute, a businesstechnology research organization. Previously, he was an Affiliate Professor and Director of the Center for the Innovative Application of Technology with the School of Technology and Information Management at Washington University. He is an internationally recognized



expert on technology transfer, software engineering, information architecture, and data warehousing. Mr. Orr has more than 30 years’ experience in analysis, design, project management, technology planning, and management consulting. He is the author of Structured Systems Development, Structured Requirements Definition, and The One Minute Methodology. He can be reached at

application are installed over time. Flash cutovers are considered extremely risky, especially on VLPs.


reported in, “Royal Bank of Canada says it will be working over the weekend to update customer accounts following a computer glitch that has caused payroll delays for thousands of Canadian workers. The bank fell behind in processing salary deposits for Ontario government workers after running into systems problems during a software upgrade earlier this week.” (See [8].)


a clue concerning what’s wrong: when people believe that just because the project is important it can fail, they sow the seeds of their own destruction.

occurs when an airworthy aircraft under the control of a pilot is inadvertently flown into terrain, water, or an obstacle with inadequate awareness on the part of the pilot of the impending disaster. is a serious challenge to agile development methods, and incidentally, few people with hands-on experience use agile methods on VLPs. It is particularly risky to attempt a VLP with agile in an organization that has only inhouse experience with agile in smaller projects.


again, agile development advocates must be careful. A considerable amount of data indicates that agile development produces better products in shorter time frames on small and mediumsized projects. Because agile is so new, there isn’t much data on VLPs using agile. It is important to be sure of what you’re doing before you attempt it.

8This 3I

was a member of the NAS Trilogy review team.


a “flash cutover,” an organization installs a new application throughout the organization all at once. By contrast, in a “phased implementation,” pieces of the


agile development has it right. Test early, test often, show the user everything.

10I’ve 5CFIT




is where agile has a lot to teach. To paraphrase Cutter Business Technology Council Fellow Jim Highsmith, change early and change often until you and the user come to an agreement on what the system must do.

noticed that, in many cases, a struggling project simply disappears from management’s radar: even though the project is not finished, it is marked “ready for operation.” At this point, the time code on the project schedule changes from “development” to “maintenance.” From then on, sometimes for years, additional costs are hidden in the massive maintenance budget. And as any experienced manager knows, you can “maintain” something forever.


in many body structures, including the skin, joints, and muscles, the somatosensory sensors provide important equilibrium information as they respond to pressure and movement. The sensations enable you to know the relative position of your arms, legs, and body. This system is commonly called the ‘seat-of-the-pants’ sense because early pilots believed that they could determine which direction was down by analyzing that portions of their bodies were subject to the greatest amount of pressure. On the ground, the pull of gravity squeezes the pressure sensors in various positions of the body, indicating in which direction the earth lies. But in flight, centrifugal forces, combined with the pull of gravity,

VOL. 5, NO. 7

22 result in G-forces that make the seat-of-the-pants unreliable as an attitude indicator” [10]. 12Visual

flight rules refer to a set of regulations under which a pilot may operate when weather conditions meet certain minimum requirements. The requirements are designed to promote sufficient visibility.


the inputs takes more time than anyone expects because the design team must guess what the inputs will be used for. Unfortunately, not until you have defined the primary outputs with some specificity do the “true” inputs become clear.


using computer-aided design and drafting (CADD) systems, architects routinely develop computer models that allow users, developers, and others to “walk through” the finished structure.

VOL. 5, NO. 7

AGILE PROJECT MANAGEMENT ADVISORY SERVICE REFERENCES 1. Jones, Capers. “Why Flawed Software Projects Are Not Cancelled in Time.” Cutter IT Journal, Vol. 16, No. 12, December 2003. 2. McGroddy, James C., and Herbert S. Lin (eds.). A Review of the FBI’s Trilogy Information Technology Modernization Program. Committee on the FBI’s Trilogy Information Technology Modernization Program, National Research Council. National Academies Press, 2004. 3. Orr, Ken. “Five Big Questions.” Cutter Consortium Agile Project Management E-Mail Advisor, 4 December 2003. 4. Orr, Ken. “Answering Big Questions 1-3.” Cutter Consortium Agile Project Management E-Mail Advisor, 30 December 2003. 5. Orr, Ken. “The Last of the Five Big Questions.” Cutter Consortium Agile Project Management E-Mail Advisor, 29 January 2004.

6. Orr, Ken. Structured Systems Development. Yourdon Press, 1977. 7. Petroski, Henry. To Engineer Is Human: The Role of Failure in Successful Design. St. Martin’s Press, 1985. 8. “Royal Bank of Canada Computer Glitch Causes Payments Chaos.” 4 June 2004 ( id=11948). 9. Sobel, Dava. Longitude: The True Story of a Lone Genius Who Solved the Greatest Scientific Problem of His Time. Penguin, 1996. 10. US Air Force. “Flying Operations, Instrument Flight Procedures.” Air Force Manual 11-217, Vol. 2, 6 August 1998. 11. Varon, Elana. “For the IRS There’s No EZ Fix.” CIO, 1 April 2004.


> Agile Project Management Advisory Service

of published issues

Upcoming Topics 

Integrating Usability into the Spiral Lifecycle by Jonathan Addleston Agile Project Leadership: What It Takes to Succeed by Doug DeCarlo Software Project Metrics by Khaled El Emam

Executive Reports Vol. 5, No. 7

Pushing the Envelope: Managing Very Large Projects by Ken Orr

Vol. 5, No. 6

The Usability Challenge by Larry L. Constantine

Vol. 5, No. 5

Adaptive Components: Software Engineering’s Ace in the Hole by Paul G. Bassett

Vol. 5, No. 4

Value-Driven Project Management and Software Development by Ken Schwaber

Vol. 5, No. 3

The Software Productivity Imperative by Mary Poppendieck

Vol. 5, No. 2

Practicing Agile Requirements Management by Sam Bayer

Vol. 5, No. 1

Evaluating ROI from Software Quality by Khaled El Emam

Vol. 4, No. 12 Planning for Risk in IT Projects by Kerry F. Gentry This index includes Agile Project Management Executive Reports and Executive Updates that have been recently published, plus upcoming Executive Report topics. Reports that have already been published are available electronically in the Online

Vol. 4, No. 11 Finding Success in Small Software Projects by Khaled El Emam Vol. 4, No. 10 Building the Emotionally Intelligent Agile Manager by David R. Caruso Vol. 4, No. 9

Evolutionary Database Management: Skills for Agile DBAs by Scott W. Ambler

Vol. 4, No. 8

Lean Development and the Predictability Paradox by Mary Poppendieck

Vol. 4, No. 7

The Software Testing Landscape by Dave Higgins

Vol. 4, No. 6

Challenging the Fundamental Notions of Software Development by Robert N. Charette

Resource Center. The Resource Center includes the entire Agile Project Management Advisory Service archives plus

Executive Updates Vol. 5, No. 14

The Software Modeling Revival: How to Avoid Being Stranded in the Desert by E.M. Bennatan

Vol. 5, No. 13

The Path to “Healthy” Projects by David Hussman

Vol. 5, No. 12

Why Teamwork Remains Hit or Miss by Christopher M. Avery

Vol. 5, No. 11

Software Teams — Your Most Important Asset: Part III: Evolve or Dissolve by E.M. Bennatan

additional articles authored by Cutter Consortium Senior Consultants on the topic of project management.

For information on beginning a subscription

Vol. 5, No. 10

Paths and Practices for Agile Coaches by David Hussman

or upgrading your current

Vol. 5, No. 9

Software Teams — Your Most Important Asset: Part II: What Movitates Them? by E.M. Bennatan

to the Online Resource

Vol. 5, No. 8

Commitment Management and CMMI by David Constant

Center, contact your account

Vol. 5, No. 7

Software Teams — Your Most Important Asset: Part I: The “Make or Break” of a Project by E.M. Bennatan

call +1 781 648 8700 or send

Vol. 5, No. 6

Frameworks and YAGNI (You Aren’t Gonna Need It) by Dave Rooney

e-mail to

Vol. 5, No. 5

A Fresh Look at Software Design: Part III — Why Aren’t We Doing a Better Job? by E.M. Bennatan

Vol. 5, No. 4

Agile Practices for Global Software Development by Daniel J. Paulish and Roman Pichler

Vol. 5, No. 3

A Fresh Look at Software Design: Part II — Is It an Engineering Discipline? by E.M. Bennatan

Vol. 5, No. 2

Trends: Looking Back at 2003 and Ahead in 2004 — An Interview with Scott W. Ambler, E.M. Bennatan, and Robert N. Charette by Cutter Consortium

Vol. 5, No. 1

A Fresh Look at Software Design: Part I — Dealing with Change by E.M. Bennatan

subscription to include access

representative directly or


Workshops Agile Software Development and Project Management

Improve Your Software Development and Project Management Practices Cutter Consortium’s Agile Software Development and Project Management workshops are designed to help organizations rethink traditional software and project management so they can create a methodology suitable to their needs, tailor it to their specific projects and get the speed and flexibility required to ensure their enterprise’s competitive edge. The workshops can be customized to include client-specific examples. Workshops vary in length from one to five days to meet all of your training needs. Accessing Cutter’s experts gives you the confidence that comes from relying on the best minds in the industry. Call your account manager today to discuss a workshop that will ensure your enterprise’s project management success.

Agile Project Management: Innovation in Action Presenter: Jim Highsmith. This workshop will help you understand advanced techniques for iteration planning, collaboration and project management targeted at situations in which projects have high-exploration factors, where customer responsiveness is paramount and in which projects operate within organizations with innovative cultures. This workshop will help you encourage dramatic improvements in your project success rate and your project teams’ ability to cope with change, through the application of a unique agile project management (APM) lifecycle framework. You’ll learn when to apply APM over traditional project management and why a wellthought-out approach to APM can help you increase innovation, keep costs down and shorten your product development cycle.

Agile Requirements: Systems Visualization in the 21st Century

Cutter Consortium 37 Broadway Suite 1 Arlington, MA 02474, USA Tel: +1 781 648 8700 Fax: +1 781 648 1950 Web: E-mail:

Presenter: Ken Orr. In this workshop, Ken Orr demonstrates how business modelers and software developers can work with their customers in real time to produce new business systems models that users can see, use and test. Learn how your organization can model its businesses and prototype these models using the Next Practice integrated business modeling and systems requirements methodology. You’ll learn about the Next Practice Methodology (NP/M) and advanced modeling tool set (Next Practice Development Environment [NP/DE]) that makes it possible to visualize systems requirements the same way that architects and engineers visualize

buildings using CAD/CAM tools. You’ll learn how to illustrate high-level business context and workflow diagrams, database designs and working prototypes using the NP/M integrated approach, which is based on “time-boxed management” and targets six- to 12-week intensive requirements projects.

Deadline-Driven Project Estimation Presenter: Michael Mah. Imposed deadlines are the norm for technology projects. Yet the nature of software projects demands that teams deal with the constant dynamics of change. This presentation addresses why software projects are different than other classes of work and how the R&D “laws” of lifecycle dynamics can be used to avert disaster. It discusses benchmarking against “the competition” and addresses laws of cause and effect, so managers can negotiate viable commitments using proven techniques for software project estimation. It reveals how to use productivity baselines for estimation, critical flaws in “traditional” planning processes, risk management techniques for fixed deadlines and what to do about “dangerous metrics.”

The Extreme Programming Workshop Presenter: Joshua Kerievsky. If you’re looking for a workshop that will give you a thorough education in Extreme Programming (XP), taught by people who coach and program on real XP projects, you’ve found it. XP takes best practices to the extreme in order to produce highly productive, agile and confident software teams. To fully appreciate XP, you have

Agile Software Development and Project Management

to do XP — not just a few of the XP practices, but all of them, together. In this five-day workshop, you will learn XP by doing XP on a simulated XP project, in a simulated XP environment, complete with customer and developer staff and experienced XP practitioners. You will understand why pair programming works after you’ve done it with multiple partners working on diverse tasks in this workshop. You will understand how XP simplifies requirements during planning sessions. And you will discover the power and importance of test-first design from experiencing the simplicity that results from following this best practice.

Implementing Lean Software Development Presenter: Mary Poppendieck. Manufacturing and service companies use lean techniques to improve quality while decreasing cost and delivery time. Why not apply these principles to software development? This workshop helps you identify and eliminate the real waste of software development. You learn how to apply the important lean techniques of customer focus, process flow and data-based decisions to your software development process; ensure your software development process delivers real customer value; increase quality and decrease cost through the effective use of feedback; reduce complexity and manage risk; and develop an implementation plan that leverages your company’s lean or lean Six Sigma initiative.

Leading Successful Projects Presenter(s): Tom DeMarco and/or Tim Lister. Today, a project needs to be run as a spry organism, nimble to the changes it’s sure to encounter. We think of ourselves as systems designers, but the project is a system, too — and do we ever turn our skills to properly designing it? All of the heuristics governing system design can profitably be applied to design of the project. Tom DeMarco and Tim Lister turn their attention to proper design and implementation of software projects. The goal is to help your projects be more productive

and better able to turn out quality results. This workshop will prepare you for a leadership role in designing, populating and inhabiting these adaptive project organisms. You’ll get practical guidance to help your project meet its specific challenges and to achieve its promise of success.

Risk Management for Software: Learning to Contain, Mitigate and Manage the Uncertainties of Software Development Presenters: Tim Lister and Tom DeMarco. Projects that deliver real benefit are full of risk. We have to develop ways to discover the lurking risks, estimate their impact, optimize our response and monitor for change. This workshop helps you identify and quantify the uncertainties that threaten software development success. For each uncertainty, you learn to contain, mitigate or eliminate its impact. You explore the link between risk and opportunity, the role of the postmortem, how to separate resultant and root causal risks, the roles during the risk discovery process, how to perform backward root cause analysis, a modified syntax of risk declaration and more.

Testing and Refactoring Presenter: Joshua Kerievsky. Learn just how brilliantly effective test-driven development (TDD) can be, how to write good test code and how to practice the rhythm of refactoring. Although its name implies that TDD is mostly about testing, it is primarily about design: it keeps you focused on exactly what you need to build and helps you avoid over-engineering. Much of what you’ll do in this workshop is write code. When you’re not writing code, you’ll be studying and implementing refactorings, gaining experience with TDD, participating in interactive demos (such as the refactoring fishbowl) and learning about automated testing techniques. This workshop is taught using Java or .NET and can be customized to meet your specific needs.

Workshops You Can Become Agile in One Day: Scrum Overview and Application Presenter: Ken Schwaber. If your organization is uncertain where a project is, Scrum provides a way to have greater visibility into it. If you have a project that needs to be turned around right now, Scrum offers an easy, low-risk way to do it. In this one-day workshop, you get an overview of Scrum theory, practices and actuality so that you’ll have an understanding of Scrum, why it works and how it works. You’ll discover the benefits one company realized by studying how that company applied Scrum processes to a complex project. You’ll have the opportunity to discuss how Scrum could help your organization and you’ll discover how Scrum and its practices can be applied to one of your specific projects.

Other Agile Software Development and Project Management Workshops:  Agile Software Development: A Review of Agile Methodologies with Jim Highsmith  Asset-Centric Software Development for Senior Managers with Paul G. Bassett  Extreme Project Management Masterclass with Rob Thomsett  The Extreme Project Management Workshop with Doug DeCarlo  Managing the Deadline: A Project Management Masterclass with Suzanne Robertson  Mastering the Requirements Process with Suzanne Robertson  Rapid Software Testing with James Bach  Scrum Project Quick Start with Ken Schwaber  Software Estimation: A Wolf in Sheep’s Clothing with E.M. Bennatan  Software Project Management with Johanna Rothman  A Taste of Extreme Programming with Kent Beck

For more information, call +1 781 648 8700 or visit

About the Practice

Agile Project Management Practice Cutter Consortium’s Agile Project Management Practice helps companies succeed under the pressures of this highly turbulent economy. The practice is unique in that its Senior Consultants — who write the reports and analyses for the information service component of this practice and do the consulting and mentoring — are the people who’ve developed the groundbreaking practices of the Agile Methodology movement. The Agile Project Management Practice also considers the more traditional processes and methodologies to help companies decide what will work best for specific projects or teams. Through the subscription-based publications and the consulting, mentoring, and training the Agile Project Management Practice offers, clients get insight into Agile methodologies, including Adaptive Software Development, Extreme Programming, Dynamic Systems Development Method, and Lean Development; the peopleware issues of managing high-profile projects; advice on how to elicit adequate requirements and managing changing requirements; productivity benchmarking; the conflict that inevitably arises within high-visibility initiatives; issues associated with globally disbursed software teams; and more. Products and Services Available from the Agile Project Management Practice

• • • • •

The Agile Project Management Advisory Service Consulting Inhouse Workshops Mentoring Research Reports

Other Cutter Consortium Practices

Cutter Consortium aligns its products and services into the nine practice areas below. Each of these practices includes a subscription-based periodical service, plus consulting and training services.

• • • • • • • • •

Agile Project Management Business Intelligence Business-IT Strategies Business Technology Trends and Impacts Enterprise Architecture IT Management Measurement and Benchmarking Strategies Risk Management and Security Sourcing and Vendor Relationships

Senior Consultant Team The Cutter Consortium Agile Project Management Senior Consultant Team includes many of the trailblazers in the project management/peopleware field, from those who’ve written the textbooks that continue to crystallize the issues of hiring, retaining, and motivating software professionals, to those who’ve developed today’s hottest Agile methodologies. You’ll get sound advice and cutting-edge tips, as well as case studies and data analysis from best-in-class experts. This brain trust includes: • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • •

Jim Highsmith, Director Scott W. Ambler James Bach Paul G. Bassett Sam Bayer Kent Beck E.M. Bennatan Tom Bragg David Caruso Robert N. Charette Alistair Cockburn Doug DeCarlo Tom DeMarco Khaled El Emam Kerry Gentry Ron Jeffries Joshua Kerievsky Brian Lawrence Tim Lister Michael C. Mah Lynne Nix Ken Orr Mary Poppendieck Roger Pressman James Robertson Suzanne Robertson Alexandre Rodrigues Johanna Rothman Lou Russell Ken Schwaber Rob Thomsett Colin Tully Richard Zultner

Pushing the Envelope  

The report discusses why big software project fail and the role that Architecture should play in such project.