Managing Software for Growth: Without Fear, Control, and the Manufacturing Mindset

Nav:Home > Textbooks > Computer Science > Managing Software for Growth: Without Fear, Control, and the Manufacturing Mindset

Press:Addison Wesley Addison-Wesley Professional (July 22, 2003)
Publication Date:2004
ISBN:9780321117434
Author Name:Roy Miller
Pages:174
Language:English

Content

Business managers -- whether consciously or unconsciously -- view software as a manufacturing exercise. 
They force software project managers to predict the unpredictable accurately.
This strategy has a fatal flaw: it doesn't work.
Managers need to replace it with an approach that allows their teams to end up with software real people want when the project is over, not what those people thought they wanted at the beginning.
This book presents a realistic strategy for creating great software when requirements start off ambiguous and become clear only as teams moves through the fog.
The author teaches readers to cede control of key elements to the development team, thereby helping the team produce more robust applications in a more timely fashion.

About the Author

ROY MILLER has spent over 10 years in IT, beginning with several large projects as a team lead and project manager at Andersen Consulting (now Accenture), and most recently as a team lead, manager, and XP Coach at RoleModel Software. 
During that decade, he kept telling himself, "There has to be a more realistic way to create software people want and need." Managing Software for Growth is the result of his desire to prove it.
Miller is the author of numerous articles and papers, and is the co-author of Extreme Programming Applied (Addison-Wesley), which he wrote with Ken Auer in 2002.

Excerpt. © Reprinted by permission. All rights reserved.

A large majority of software development projects still cost too much, take too long, and don't deliver what they promised at the beginning. 
That's if they actually deliver something at all.
A significant number get cancelled before there's any useful software to hand over.Not all software development projects fail, but it certainly can seem that way.
In 2000, one industry survey claimed that roughly half of software projects were over budget, late, or didn't deliver what they promised.
Another roughly twenty-five percent were cancelled before delivering much of anything at all.1 These numbers were an improvement over the figures for the first survey in 1994, where the results were abysmal, but this is hardly cause for celebration.
Almost three quarters of software projects still produce junk, late, for a lot of money, and at the expense of people involved.If you ask industry observers and corporate IT leaders why this is so, I'll bet you'll hear a standard response.
They'll tell you projects with more executive sponsorship, smaller scope, greater user involvement, firmer requirements, better estimates, better programmers, and more solid project management are more likely to succeed.
Projects that deviate from these things are proportionately more likely to fail.These are good things, but this recipe for success isn't very satisfying.
It's difficult, if not impossible, to have all of these things on a single project at once.
Most projects can't, and don't, but they have to create software anyway.
Even more disturbing is that some projects have all these things, at least according to the checklist, and still fail.
Why?Why are business leaders willing to spend millions of dollars each year on sub-par products from IT, or from services firms they hire? How come they're willing to spend just as much money, and more, on advice from IT management gurus whose advice often doesn't work, or is impossible for real human beings to implement?The Manufacturing MindsetSoftware development in most big companies, and even in many smaller ones, is based on a single, unquestioned metaphor:We manufacture software.The Industrial Revolution created a psychology for work and management.
People who lived during that time, especially in the first two-thirds of the twentieth century, thought about most aspects of life in terms of industrial production.
Business leaders did the same.
This was natural.
Software development for mission-critical business systems began to gain serious momentum in the 1950s, in the heyday of the Industrial Age.
It isn't surprising that software developers of that time, and the business leaders who employed them, thought about software development like they thought about other kinds of "production." In fact, traditional explanations about why software projects fail, and traditional methods of avoiding failure, assume that software development is a control and optimization problem, much like factory production.
These explanations have held sway for almost fifty years.How would you increase quality and efficiency in a factory? You would make things predictable.
You would break work down into very small steps, and have workers do the same thing all the time.
You would make sure to engineer what you're producing to the last detail, then you would determine the optimal production steps and schedule.
If you assume software development is like a factory, and software developers are like workers on an assembly line, these success factors make sense.
Unfortunately, the assumption—often unconscious—that software development is an efficiency optimization problem has led IT organizations to a dead end.Bad AssumptionsThe manufacturing mindset is familiar and comfortable for people born in the twentieth century.
It's reassuring for us to believe that we can extend that mindset to software development, but this belief rests on three bad assumptions—assumptions that are reasonable in a manufacturing context but flawed in a software development context.
These are:That we can make useful, detailed, accurate long-term predictionsThat we can control people and events to make our predictions come trueThat our attempts to control will produce the results we wantAll of these assumptions turn out to be wrong most of the time.We can only make useful, detailed, accurate long-term predictions if changes we make have proportional effects, and if all the pieces of the problem are independent.
If changes can have disproportionate effects, or if the pieces of the problem are interdependent, then accurate, detailed, long-term prediction becomes impossible.Suppose I tell you that the Design phase of a project will be done on a specific date.
What if the design isn't done? Then the Design phase isn't really done.
Or what happens if the team encounters a problem in the Build phase or the Test phase, and they have to redesign a certain component completely? Then not only is the Design phase not done when I said it would be, but it also isn't done when we thought it actually was.
And the component won't look like my prediction said it would either.
The bad news is that these things always happen for systems that do much of anything interesting.
When we plan at the level of detail most people are used to, we're trying to predict specific details about the long-term future of something affected by a dizzying number of variables that relate to each other in complex ways we can't sort out completely ahead of time.Even if we could predict accurately, though, we wouldn't be able to control everything like we want to.
Real-life problems defy control.
There are too many variables interacting in ways too complicated for us to understand completely.
In common language, this is known as the Law of Unintended Consequences.
The slightest change, a seemingly innocuous little modification, can have disproportionate effects that we can't predict.
We have very little hope of controlling much of anything really complex.This gives already overstressed managers intense heartburn, which leads them to obey what Tom DeMarco calls the First Law of Bad Management:If something isn't working, do more of it.When things seem out of control, most managers react understandably—they try to control more.
When they do that, they unintentionally squeeze out the potential for novel discovery their teams need in order to solve problems they've never solved before, which is what software development is all about.
Managers bent on control force their teams to follow all the checklists for their industry certifications (choose your flavor) right into the ground.
The best control-oriented managers can hope for is that their teams will produce the software people thought they wanted when the project started, not the software people end up wanting after they learned along the way.
That's not the result anybody wants.These three bad assumptions—that accurate, detailed long-term prediction is possible; that control will make our predictions come true; and that control will give us the results we want—explain why software projects based on the manufacturing mindset so often end up in the weeds.
What fascinates me is that, at least at some level, we know this ahead of time.
We know we can't predict the future very well, but we act as if we can, and we take our attempts very seriously.
The reason is that we forget we're all human, and we've been conditioned to think in certain ways by our experiences, by what we've been taught, and by the times we live in.
When managers think and behave according to their conditioned mindset, they unconsciously lock their teams into the status quo of failure, frustration, and disappointment.A New Way of ThinkingThe manufacturing mindset might have made sense back when software development was a simpler problem.
We're past that point, and the analogy doesn't work anymore.
Software development today is a completely different animal.
It's a creative process, not entirely mechanistic.
Software emerges as we create it, sort of like a painting.
What managers need is a new way of thinking and behaving which recognizes that software development is a creative process, but which still allows them to act intentionally and to make realistic plans.
I call this new way of thinking growing software.This book is about applying a principle that seems too simple to most people.
It is that we should grow software, not build it.
Software development is a messy, creative problem that is virtually impossible to predict or control in the way we think it ought to be.
If that's true, then our traditional approaches to planning and running software development projects make very little sense.
Instead, software development should look more like human beings feeling their way toward solutions they can't quite put their finger on until they see them.
I'm sure your own experience tells you this is not how most organizations develop software.When human beings try to solve messy problems like software development, they often act like what complexity theorists call a complex adaptive system (CAS).
These systems share three important characteristics.First, complex systems are based on networks of agents acting in parallel.
In general, these agents can be any components in a system, such as cells in your body or amoeba on a slide.
The really interesting systems, though, are the ones with human agents.
Human agents are not always rational, they use strategy in their interactions, they have to operate on incomplete information, and they have previous experience to influence their current behavior.
The programmers, managers, sponsors, and other people involved with software projects are the agents of the software development CAS.Second, the number of possible actions each agent can take at any point is huge.
As a system gets more complex, the space of possibilities for what happens next is vast.
The best human agents in such systems can do is to use rough rules of thumb, built into collective memory, or history based on feedback from trial and error.
These skills become the predictive model agents use to decide what to do next, without any centralized control.
Agents "explore" their way to workable solutions.
This is what the programmers, managers, and other people involved with software projects do, even when managers try to force things to go according to plan.Third, these imperfectly intelligent agents exploring their way to workable solutions are subject to conflicting constraints.
They can't simply do whatever they want.
Most of the time, they have to try to make many different stakeholders happy.
In real life, it's virtually impossible to make everybody happy at the same time.
The best agents can do is pick their battles and do the best they can.
Software development projects are like this.
Programmers, managers, and sponsors have to balance the forces as best they can, and still try to deliver the "right" solution.These three characteristics—imperfectly intelligent agents, exploring a vast space of possibilities, while trying to balance conflicting constraints—are the same ones that define what complexity science calls a CAS.
In biology, in economics, in public policy, and in software development, the complex systems that survive and thrive behave like CAS.
In human terms, organizations that survive and thrive do the same, allowing people to solve messy problems in ways that make sense, which often means muddling through a vast sea of incomplete information.
Traditional software development approaches tend to ignore that.
That's why a large proportion of projects fail.
Managers and others in organizations need a new way to think about software development that respects the true nature of the problem.
That's what growing software is.Growing software allows us to get the software we want at the end of projects, not what we thought we wanted when we started.
When we start a project, we almost never know exactly what we'll end up wanting.
We should stop pretending that we will.
The only way that can happen is if we recognize that software development won't produce the results we want unless we think about it differently.Growing software requires a significant change in the way managers at all levels (and developers, for that matter) think about software development.
It also requires different behavior.
Old habits die hard, and we have half a century of conditioning to overcome.
I believe it's possible, but the change can be painful.
As is so often true of life, the road to better goes right through uncomfortable.
But I don't believe it's hopeless if we're committed to making things better.Implications for Software DevelopmentThe corporate IT world can become a radically different place if people involved with developing software can change the way they think about it.
That means people who write software (programmers), people who buy software (project sponsors), and people who manage software projects (managers).
This can be particularly difficult for managers, who are often under crushing pressure from above.If software development is a complex adaptive system, we cannot think of the people involved with creating it as cogs in a machine, or as workers on an assembly line.
They are more than that.
They are agents, albeit ones with perpetually limited information.
They are interacting with other agents to create an emergent reality that is greater than the sum of the parts.
Managers need to see them that way, especially since managers themselves are agents in the system as well.
Those same managers need to participate in the process, rather than trying to control everything, which is much different from what they're taught and conditioned to do.What we will see is rather startling.
The act of creating software to meet a specific business need involves agents with limited information exploring their way through a vast space of possibilities by following simple rules of action and interaction.
That complex and unscripted dance, which often appears chaotic, can produce the kind of order that makes it possible to solve messy problems like software development.This BookManaging Software for Growth is for managers.
That describes everybody in the management hierarchy, up to the CEO.
But I wrote this book for people sandwiched in the middle of their organizations, the people who most often receive demands from above and somehow have to produce the demanded results.
What you learned in school and on the job, the thinking you internalized, and the realities of software development make it tremendously difficult for you to make that happen.
I believe it can happen, but you have to think and behave differently.I organized this book into four parts:Part I Manufacturing Software.
What it means for projects, why we keep doing it, and why it makes very little sense anymore.Part II Complexity.
How complexity science can help us understand software development, and be more realistic about it.Part III Growing Software.
What it is, how it's different from manufacturing software, and what modern software development should look like.Part IV Change.
What managers should do differently, starting tomorrow, to help their organizations grow software.My hope is that reading this book will help youProduce results that delight the people in your organizationDo that without having to predict things you can't predict, or control things you can't controlParticipate in creating a more realistic, enjoyable place to workThe point of talking about this at all is to answer one fundamental question: How can you realistically help your teams create software that people end up wanting? Writing software is difficult.
Writing great software is harder still.
Consistently writing software that makes customers cackle with glee seems like a dream to most managers and programmers.
I believe this is possible, if everyone involved is solving the right kind of problem realistically.
That's what Managing Software for Growth is about.

Tags

Textbooks,Computer Science,Software Design & Engineering,Computers & Technology,Programming,Software Design, Testing & Engineering,Software Development,Software



 PDF Download And Online Read: Managing Software for Growth: Without Fear, Control, and the Manufacturing Mindset



Comment

 
 

Comment List (Total:9)

  •     There are some things in this book that are very good, but the main punchline fails to deliver. I'll highlight the ups and downs below.
  •     This book starts by describing how Taylor's Scientific Management has lead us down the incorrect road of thinking software can be managed like steel manufacturing. The author does a great job of presenting these fundamentals and presenting the conclusion that a new model for thinking about software is needed.To find an alternative model the author draws on ideas from complexity science. He presents the idea that we need to think of software as growing rather than something we control if we can just add one more level of detail to our Gantt charts. If we think of software as emerging from the environment in which it is started we can exert influence on the growth of the software so that it grows toward our goals. However, we still can't control it like we can control steel manufacturing or an assembly line.An earlier reviewer has commented that there is much to disagree with in this book. I think the author of the book would scream "Yes! Exactly!" Of course there's much to argue and debate with in this book. The book is full of original thinking and is proposing a radical shift in the thinking of most software professionals; that is bound to stir up debate and some disagreement.
  •     Expertly written by Roy Miller (a team lead and XP coach at RoleModel Software with over 10 years of information technology experience), Managing Software For Growth: Without...
  •     Before adding my two cents, I must state that I previously co-authored a book with Roy Miller and have worked with him for almost three years.
  •     I read thru early review drafts of this book made available from the yahoo-group website for the book. What I liked most about it is that it was able to successfully convey to me not only the "manufacturing mind-set" it describes (in contrast to agility), but it also conveyed the foundations of that mindset and why it is held by so many in the industry. I never really had a good enough appreciation for that until reading this book.I think the book also does the same "justice" to the agile-mindset for "growing" software thru "emergence". It describes the foundations of agility and emergence, where they come from, what they mean, and what evidence there is and isnt for their validity.The book touches on elements of Taylorist management theory, as well as complexity theory in an approachable manner that doesnt require an advanced degree in math or physics (which is a plus :-).I also particularly like the annotated biliography at the end of the book. It lists more than just the titles of the "giants" the author stood upon to write this one, it excerpts and distills the key central concepts from each one that are relevant to the book.
  •     Complexity Science has been applied previously to software management by (at least) two leaders in software develoment: Jim Highsmith and Mike Beedle.
  •     This is THE book I recommend to clients who are hesitant to use an agile approach in developing software. It's an excellent book for both managers and developers.
  •     One of the oldest dilemmas in software development has been "Is software development art or science?". Of course, it's neither, but the science and engineering mindset has generally held sway, and deviations from this approach have been treated as aberrations, to be eliminated so that we could be more 'scientific'. But what if that's the wrong direction?Roy Miller explores the origins of the scientific/engineering/manufacturing mindset in software engineering, and then goes on to suggest reasons why this metaphor might not be appropriate for modern software development. He weaves strands from a number of disparate disciplines, including biology and chaos theory, to suggest alternative approaches to software development. Multi-disciplinary approaches are a common theme in modern scientific research, and it's good to see someone bringing these influences to software development.Roy's fundamental message (and he of course justifies this in detail), is that if you're looking for predictability in software development, you better get over it. Software development is a messy, human endeavour, and if we keep doing what we did, we'll keep getting what we got. Roy suggests alternative metaphors and approaches that might be more succesful in growing software.I see that at least one other reviewer has commented that the Taylorist manufacturing mindset can be applied successfully to software development. Even if that's true, it doesn't make it the only game in town, and it's worth reading Miller's work for an alternative. My own experience suggests that the Taylorist manufacturing model only works when a software project is large enough to make individual human differences irrelevant, and that projects of that size are becoming less common, and generally fail anyway.Read this book, even it it makes you uncomfortable or doesn't change your mind. You'll gain at least one new idea, and that's worth the price of admission.
  •     This is a very important work that helps the software industry to take much needed steps to gain reliability and authority again.

Recommand Books

 

Legal Theory & Systems,Cooking Methods,Physics,Dentistry,Other Team Sports,Infantil y juvenil,Cars, Trains & Things That Go,Christian Denominations & Sects Book,。 FreeBook 

FreeBook @ 2018