When people talk about software process improvement (SPI), they usually mean a specific methodology for improvement, such as the Capability Maturity Model (CMM), or its big brother, Capability Maturity Model Integration (CMMI). These models address what to do, not how to do itwhether that's creating documents, getting signoff, review, audits, or other practices. Invariably, these new practices add new kinds of work, meaning that an already stressed team has to do more in order to ship software products.
When I think of software process improvement (SPI), one word comes to mind: pain. The organization is probably experiencing pain already; that's why they're pursuing SPI. Perhaps they need some certification to win a contract; maybe the software is buggy; or possibly they can't accurately predict schedules or control scope; for whatever reason, the team is already in pain.
But There's More!
Notice that we're not talking about methods of doing, but rather methods of improving. Your organization doesn't just get to "adopt" CMM or CMMI; instead, you have to create your own implementation that meets or exceeds the requirements of the process improvement method.
In order to do that, you have to know what your gaps are, so your team has to do an assessment and gap analysis. Once the gap analysis is done, you get to create an implementation plan. Once the implementation plan is in place, you have to create the policies, procedures, and documents that must be followed to conform to the standard. Only then, when the policies are in place, can the team begin to follow the newly required additional practices, audits, and measurement-gathering steps in order to conform to the standards.
Like I said: Pain.
Let's call the approach that I've just described comprehensive. It's designed to gather all the relevant information about how work is done, come up with one true way that the work should be done, and then standardize or codify in a rulebook the way the work will be done.
But that's not the only way to do it. Let's take a moment to consider another approach.
A Different Kind of Process Improvement
We'll call this approach organic. Instead of gathering all the data, you look at specific projects, asking what the risks are right nowwhat's the single biggest thing we can change right now to have a positive impact?
I know. Crazy!
Notice that the organic approach looks at external ideas as just thatideas and food for thought, not standards requiring compliance. The thinking goes like this: In a way, creating a standard that applies to everyone is essentially prescribing that same solution to everyonewhether or not they have the problem.
In the medical field, we don't call that "best practice"; we call it malpractice. We need to be careful about offering solutions without first examining the patient.
In software development, the business itself is the patient.
We might not prescribe what to do, but it can be helpful to suggest ways to analyze, to diagnose, to decide what to do next. I'll discuss two:
- Management by walking around and listening
- Risk analysis and critical success factors
Management by Walking Around (and Listening)
Years ago, I was a contributor on a software project that was running under stress. In one meeting, the development team expressed skepticism over the project dates, but by the end of the meeting, the manager had cajoled them into agreement.
A few minutes later, in the bathroom, I overheard people saying what they really thought. There was no way that they were going to hit the date, everyone knew it, and they all felt that they had transferred the responsibility to the line manager.
Somehow I doubt that he felt that way. That guy should have been in the bathroom.
Believe it or not, the idea of "management by walking around" (MBWA) is a tried-and-true management technique, discussed in books like Mike Cohn's Succeeding with Agile: Software Development Using Scrum and Karen Otazo's The Truth About Being a Leader. The idea is trivial: Managers can see opportunities for learning and improvement every single day, if they just pay attention and listen. So, managers, get out of the office and see the work.
But there's more to MBWA than just walking around; that "and listening" part is important. If the line manager had been practicing it all those years ago, he would have found out about the reality of the project, and perhaps he would have been able to steer toward reality, instead of steering toward wishful thinking.
In addition to listening, an effective manager needs to recognize problems, triage them into priorities, and visualize the fixes. To do this, the leader needs to have a deep, intimate understanding of the business processes.
This deep understanding of the business is central to the Japanese method known as the Toyota Production System, something that Americans have codified into the lean manufacturing and lean software development.
Lean software development has goals and rules, such as identifying bottlenecks, achieving one-piece "flow," and "pull." These goals and rules maximize system throughput, and they're helpfulas far as they go. Yet the Japanese had great success without codifying the rules; instead, they simply understood the business and found waste.
Do you see someone sitting around, waiting for a process to finish? That's waste.
Someone else is blocked, waiting for a document or decision? Waste.
A third person is creating a document only because it's required to check a box for a process audit, or writing a five-page status report that no one will read? You got it. Waste.
For the record, I'm not talking about eliminating coffee breaks or team lunches, because those events often involve sharing of information that can accelerate delivery. Instead, I'm talking about the obvious sort of insanity that everyone can see but nobody talks about.
Here's a crazy idea to get teams delivering: Get them out of their email and into the work.
How do you find opportunities for improvement? Try walking around and listening.
Critical Success Factors
Determining critical success factors (CSF) is another well-established business technique. Your company can apply CSF at any level, from the development team to the board of directors.
The idea behind CSF is deceptively simple: Identify the two to five things your team has to do in order to succeed:
- If your team doesn't accomplish their CSFs, nothing else they do matters.
- If your team accomplishes their CSFs, nothing else they fail at matters.
There's more to CSF than those two bullet points, yet they sum up the main idea.
When a team tries CSF for the first time, they typically go through a fascinating process:
- Management claims that there's really only one issue; perhaps it's getting the "SquareCalc" program done by the end of the year.
- The team finds a bunch of out-of-the box ways to get SquareCalc done. Perhaps methods like these:
- Skipping some mandatory corporate training.
- Holding off on maintenance of a different system.
- Killing a half-dozen small projects in order to "feed" SquareCalc.
The beauty of the CSF approach is that it limits youyou get five things. If the team needs more than that, it has to make some tough decisions about priorities. Using CSF won't get that priority issue solved, but it will bring the issue to the surface, and that's a good place to start.
You can also apply CSF at an organizational level. Instead of looking at projects and schedules, you'll look at key processes that lead to success or failure. For example:
- How schedules are built
- How the team controls changes in requirements at the beginning of projects
- How the team triages fixes versus enhancements at the end of projects
- How specific functions communicate and hand off work
Done in this way, CSF is very similar to those painful improvement methods I mentioned earlier, with one subtle difference: The improvement is customized to your business, addressing the things you need fixed right now, instead of trying to fit a checklist or general template of "maturity levels."
Process or Skill?
One of the main assumptions of the process-improvement movement is that you can define processthat you can create recipes for software development, just as you would describe how to make a sandwich.
Not all work is like that. Some processes, like flavoring a soup, are empirical in nature; you flavor to taste, adjusting your mix of spices, sweeteners, and so on to fit the situation. As Alistair Cockburn, coauthor of the Agile Manifesto, once put it so well in his article on characterizing people in software:
A few months later I took off my consultant's hat, put on my ethnographer's, and studied the ways in which the team was actually behaving. What I saw staggered me: The moment-to-moment process the team followed was so intricate that I couldn't possibly write it all down, and if I could, no one else could possibly follow it. It matched my process only in the most superficial way.
Every project member needs to make dozens, if not hundreds of judgment decisions in a daydecisions such as whether to make a decision or gather more information. Judgments like the best way to gather that informationby phone, email, or walking over to someone's cubicle. They also need to make decisions about who to consult on a sticky problem, or how to design an interface between two systems. The process description might mandate a meeting, but it likely won't describe how decisions are made.
A process description might help (a little), but fundamentally, no document can teach a golfer how to swing well. Instead, the muscles need to be trained through practice, repetition, and experience.
It works the same way in software development. At their heart, MBWA and CSF are methods of getting to a skill, and, more than that, identifying which skills need the most improvement right now.
What Does That Mean for Us?
The time, money, and energy we spend on improvement can always be spent on something elseperhaps actually doing the work. At the very least, the organization could keep the money in the bank, which means that process improvement has a cost. Documentation implies a cost also, but if the team wants to keep those documents updated as things change, additional cost is involveda maintenance cost.
Sure, you could pick up a methodology book, fly in a consultant, conduct a gap analysis, and do all the rest. In this article, I've tried to show you a different way: Identify the current obstacles, the current blocking conditions. Work to remove those issues, institutionalizing the fix reluctantlyand only if you're certain that the benefit outweighs the cost of yet another "required practice."
This is organic improvement; and, yes, it is possible to do it in a more comprehensive way. You might take notes on what the fixes are for several projects over a few months, and then look hard at how those fixes could be scaled up to the entire organization. The important thing is to map the fixes and improvements to real problems that the organization actually has, while making sure that the new benefit outweighs the cost.
Notice that this kind of improvement is more than just a "process" improvement, in that it includes notions like skill, hustle, and human motivation in its considerations. Organic improvement tends to focus on outcome, not process. Indeed, some experts use an entirely different term than process, calling this kind of work performance improvement.
It turns out that there is a word for this kind of critical thinking and making up your own mind, doing what you believe is right based on your own values: We call it maturity.
Let's all endeavor to be mature.