Month: May 2007

Late projects, man-months and the software crisis

~ Andrew Stellman ~ 1 Comment

I recently got a question recently asking about Fred Brooks’ famous quote: “Adding manpower to a late software project makes it later.” The person asking the question took it literally, and was asking about whether this meant that there’s no way any schedule estimation technique could indicate that adding manpower could shorten delivery time.

Fred Brooks wasn’t really talking about schedule estimation techniques, or writing hard-and-fast rules. He was trying to help software engineers break some very bad habits. And trying to fix a failing project by throwing bodies at it is one of the worst habits software engineers have. See, as it turns out, some tasks take however long they’re going to take, and throwing bodies at the problem won’t change that. Or, as Brooks wrote, “The bearing of a child takes nine months, no matter how many women are assigned.”

Business Plan

I think the best way to understand that Brooks quote was to understand why he first wrote “The Mythical Man-Month” back in 1975. At the time there was something called the software crisis. The field of software engineering, still young at the time, had run in to a serious problem: people were having a whole lot of trouble writing software on time and under budget.

By the time the software crisis was in full effect, the engineering world had already matured a great deal. The great civil engineering projects of the early 20th century taught us how to manage enormous projects effectively. (Hoover Dam was finished two years ahead of schedule! And Henry Gantt invented of everybody’s favorite scheduling tool around 1910.) And by the mid-1970s, the world of computers had matured to the point where most large corporations had computer programming teams and IBM “big iron” timesharing mainframes. It seemed like the tools were there, and the people had the skills to use them.

So why were all of our software projects so late?

Personally, I blame the person who coined the term “software”. The thinking, I believe, went like this: “Hardware is really expensive, and basically impossible to change. But luckily, we’ve got these programs that we can constantly tweak to suit our needs! That’s the opposite of hardware, which can’t be changed.”

By the time the mid-1990s came around, the field of software project management had matured to the point where we knew that change was the biggest factor in late software projects. Changing needs and requirements meant pulling apart the underpinnings of our code and caused us to have to do expensive (and often ineffective rework). Changing technology kept giving programmers “silver bullets” that they thought would make this next project run a whole lot more smoothly than the last one. And the (perceived) ability to change the software architecture and design in the middle of the project caused a lot of teams to fail to plan their projects well (or at all!).

The worst part of it was that it seemed like the software crisis never ended — there were still an enormous number of projects that were still coming in very late and over budget, or failing to complete at all. But at least by that point we knew the root cause of the problem and had tools to help us fix it. In fact, we were learning that adapting to change and getting those changes under control was an important part of effective software projects.

So how did we know that those things were causing the crisis in the first place, and that fixing those problems would get our software out the door on time and within budget? We can thank Fred Brooks for that. “The Mythical Man-Month” was, as far as I know, the most important software engineering book of its time. It laid out the real problems that caused software projects to run into trouble, and showed us a lot of very useful ideas about planning our software projects, structuring our teams, estimating the work, communicating with each other and our stakeholders, and writing our documentation. And a lot of what he wrote still resonates very well with modern software projects. (A lot of the tools and techniques that Jennifer Greene and I wrote about in Applied Software Project Management were designed specifically to address those root causes of software project problems.)

And that gets back to Brooks’ quote about adding manpower to a late project. When Brooks talked about a late software project, he was talking about a project that had constantly shifting requirements and needs, poor planning, and communications problems both among the team members and between the team and the stakeholders. In a situation like that, many project managers will panic and just try to throw bodies at the situation. (I’ve seen this many, many times — and so have almost all software engineering professionals.) But those extra people will just cause the bad project plan to become even more inaccurate, and they will compound the already serious communications problems. That, in turn, will cause the team to spend extra effort dealing with those problems. And that extra effort will often be larger than the man-hours that the newly added team member will add to the team.

That’s why Brooks called his book “The Mythical Man-Month”. If you need to get your two-person project done in six months, you can’t just add another person on and magically get it done in four months. Adding a person to a project for a month doesn’t automatically reduce the total effort by one man-month. The effort that one person can contribute to a project varies enormously based on the project circumstances. His “adding manpower” is saying that if the project is already late, then adding an extra person can actually subtract man-months from the project.

PMP Study Tip: Write Your Own Questions

~ Andrew Stellman ~ Leave a comment

The PMP exam is all about questions. That’s a little obvious, I know. But think about it for a minute: you’ve got 4 hours to answer 200 questions. It’s a nerve-wracking situation if you’re not in the habit of taking exams — and few professionals are. That’s why one of the best ways that you can prepare for the PMP exam is to write your own questions.

But don’t you need to be an expert to write your own questions? Definitely not! It turns out that there are only a few ways that make sense for presenting the kind of information you’ll see on the PMP exam. Getting a good feel for them will really help you get more comfortable with the material, and it will help you separate the ideas being tested on from the wording of the question. Most of all, it will help you think more about the material, and come up with practical situations on your own for applying it. And that’s a very effective way to study.

It’s hard to know where to start, though. It turns out that writing good questions takes practice. That’s why Jenny and I added the “Question Clinic” Lose Weight Exercises to Head First PMP. It turns out that a lot of the questions on the exam fall into just a few categories. There are different strategies for each kind of question, so recognizing the question type can help you get more answers right. But even better, being familiar with how questions work can really help keep you calm on the exam, because you’re faced with a lot fewer surprises. Pretty neat, right?

Bring Your Own Questions

Our friends at O’Reilly just put up a great tool on the Head First Labs site. It’s a fill-in form that helps you build your own questions, with templates from all of the Question Clinic Lose Weight Exercises in the book that show you the different kinds of questions. Even if you aren’t using Head First PMP to study for the exam — obviously, we hope you do! — you can still use this to help get a handle on writing your own questions. (You can also feel free to use the Head First PMP forums to ask questions and get study tips.)

And don’t forget: any time you run into a fact that you don’t know, or if you get a question wrong on a practice exam, write your own question for it. It’s a great way to get the information to stick to your brain!

Useful ideas for really large projects.

~ Andrew Stellman ~ Leave a comment

How do you manage a large project? There’s a big difference between managing a project team with, say, ten people and a team with a hundred. That shouldn’t be too surprising. But if you’re managing your first large project, it can be really daunting. Luckily, there’s been a lot of good work done on understanding large projects and how to manage them. The risks are bigger, and different, than anything you’ve seen on a smaller project. But if you pay attention to the tools that you’re using, you should be able to get a handle on things.

When you’re talking about a large project, communication is the area with the biggest potential risk. That shouldn’t be a surprise; after all, some people say that 90% of project management is communication. The larger your project, the more complex the communications can get.

Crowd 2007-04-30

Anyone who’s studied for the PMP already exam knows about this stuff. Remember that formula you had to memorize for counting lines of communication? Well, this is exactly why you need it! As you add more and more people to the project, the number of potential lines of communication goes up exponentially. And if you’re not prepared for it, you can really loseWeight Exercise track of who’s talking to who, and about what. So communications management — putting together a communications plan, keeping your project teams informed, managing your stakeholders’ expectations, and generally serving as a hub of communications — really starts to make a big difference.

Recently, I had to face a situation with increasingly complex communications. I was working with a consulting company that specialized in outsourced software engineering projects. We were assigned a large part of a much larger project, one that was bigger than any other project the client company had taken on. There was another major vendor on the project as well, also doing a large amount of work. In previous projects for this client, we normally didn’t do much work alongside other vendors. But in this case, I saw that the project was divided into several teams on our side, and several more teams in the other vendor’s group. It was clear to me that there would be a large risk of communications problems. So I worked with the project manager from the other vendor to establish communications channels between our leads and theirs (getting buy-in from the client, of course). By planning for and establishing a separate line of communication, we were able to take some early steps to mitigate communications risks. This is definitely paying off now, and the project is going relatively smoothly.

As your project gets bigger, your risk management also becomes a lot more complicated. Which means you need to take more care with how you plan and track your risks. If you don’t already keep a risk register, start. And definitely spend a lot more time talking with your team about risk planning at the beginning of the project. Scott Berkun has a lot of good advice in The Art of Project Management about brainstorming and talking to your team about ideas. All of that good stuff applies to discovering and analyzing risks. Take extra time at the start of the project to work with the team to identify risks, and figure out how to handle them. Are you going to mitigate them? Which ones do you need to accept? Can you figure out ways to transfer the risks? These are all questions you should be able to answer BEFORE the risk materializes. Revisit your risks at regular team meetings. This is one of those areas that really benefits from economies of scale — the more your team is thinking about risk management, the more potential and actual project problems you’ll avoid.

Risk mitigation

Is your project REALLY large? If you’re managing a project with 1,000 people, then you really need a solid metrics system. You’re never going to get a good handle on what your team is doing by just talking to people — there are simply too many people to talk to! That doesn’t mean you don’t communicate, of course. But you do need to reliably collect data, and then analyze that data statistically. This is what tools like run charts, control charts (or Shewhart charts) and trend analysis were developed for. Here’s a tip: If you’re playing in that arena, and if you have budget for consultants, find yourself someone who’s familiar with those tools to help you set up a system to gather and analyze project data. That sort of thing is very hard to set up once the project wheels are already turning, though, so you’ve got a limited window of opportunity to get it established.

(Inicidentally, if you haven’t read Scott Berkun’s book, you definitely should. I had the pleasure of being a technical reviewer for it, and it’s one of the best books on project management I’ve ever read.)