My wife and I recently undertook a small home renovation project. We have a two bedroom house and a baby on the way. As a result, I found myself trading in my posh office space for not so posh basement dwellings. Neither Leora, nor myself, especially liked the idea of me sitting in some dark and damp concrete corner hovering over a keyboard. So we hired a contractor to build out a proper office space in the basement. We told the contractor what we wanted. He planned it all out, coordinated with subcontractors, and met with the code enforcement officer periodically. The whole thing was finished in the course of about three weeks.
What impressed me the most about this project was how wonderfully uneventful it was. Granted, it wasn’t exactly a major construction project. It was just a room. But it involved a lot of people – insulators, framers, electrician, drywallers, and painters (my wife and I accepted this duty). All told, three subcontractors and at least ten different people contributed to the construction of this room (eleven if you count the code enforcement officer). Since each element of construction was built upon some previous element (e.g., framing could only be done after insulation, electrical after framing, etc.), it required a bit of coordination and an awareness of how one person’s work impacts the work of the next guy. Ten different people with a mix of skillsets and egos. It sounds like a recipe for frustration, but somehow it went smoothly and stayed on schedule.
Being a software guy, I immediately began comparing this experience to the software teams I’ve been involved with. I’ve worked with a variety of different teams over the years – big, small, agile, spiral, waterfall, fun, all-business, slow, nimble, etc. I’ve worked with, and learned from, some truly remarkable software developers on just about every team I’ve ever been on. And in most cases, the projects those teams worked on were delivered on-time (ballpark), on budget (mostly), and were considered successes. However, there’s only one team, maybe two, that I look back on and regard as laser focused and as well-coordinated as my basement room construction team. Is a team’s success a measure of their professionalism? Or do those successes hide the ugly truth? The truth being, of course, that many software development teams are fraught with dysfunction.
From the customer’s perspective, it doesn’t really matter. Customers only care about the sausage, not how it’s made. From the software developer’s perspective, it matters a lot. All software developers like to think of themselves as software professionals. But there’s a difference between a professional software developer and a professional software development team. A team of software professionals does not make a professional software team. Think about that for a moment. It’s an important distinction to make because it impacts the customer more than many folks realize.
Some might say the root cause of team dysfunction really boils down to two things – poor leadership and communication. I think that’s painting a picture with some rather large brush strokes. It’s an over-simplistic reduction. A lot of little things contribute to team dysfunction. And over time these things compound. Poor leadership and communication are often by-products rather than the cause.
In this article, I’d like to look at a few things that I think are important for software teams to get right. Understand that much of this is my own opinion. It’s based on my own experiences. No two teams are the same. And not everything I mention below may be appropriate for every team. I’m hopeful, however, you’ll find it relatable. And who knows? Perhaps you’ll discover a few ideas that may be useful for your own software team.
Let’s dive in.
Team Structure
There’s a growing trend among modern software development teams, especially the smaller Agile teams, for self-organization. Many take the position that if knowledge and responsibilities are collectively shared across the team, no functional area can be siloed, and anyone can contribute wherever help is needed. For some teams, this can work fantastically. For others, it’s disastrous. Whether or not it can work depends largely on the product and the team dynamic. For teams that need domain experts, this approach may not even be an option. It’s important that you structure your team appropriately for the product and not the other way around.
Many times a team’s structure is dependent upon the development methodology they choose to use. I talk specifically about methodology later on, but it’s worth mentioning here that a poor implementation, or even a poor choice, of a software development methodology often results in a poor team structure.
Look at your project carefully. Ask yourself questions such as the following. “Do we need domain experts?” “Do we already informally think of our team as being comprised of smaller teams?” “Is the scope of our project(s) appropriate for our team size?”
Teams need to be sized appropriately. If your team is too small, it’ll struggle to keep its head above water. If the team is too big, it can actually slow down under it’s own weight. Some methodologies even pose limits on team size. Don’t scale your team size for the sake of your methodology if it means that the size is inappropriate for the project.
Ensure your team is composed of folks with specializations appropriate for your product. Software development usually involves more than just writing code. Most software teams include testers, graphic designers, project managers, Scrum Masters, business analysts, product designers, etc. Even among coders, it’s not uncommon to have individuals who are experts in specific problem domains and whose time is dedicated almost exclusively to those efforts. Examples include imaging/optics, digital signal processing, chemistry, geospatial information systems, flight and aerial dynamics, advanced mathematics, etc.
Regarding your software engineers, it’s healthiest, I think, to have a good mix of experience levels. One of my first jobs out of college was with a company whose entire software development staff was composed of kids fresh out of college. It was a small company with a tight budget. Since junior level developers were cheap and eager to prove themselves, it seemed like a no-brainer for the company to staff-up on people like me. The problem was, of course, we had no technical leadership. There was little to no design. Even the technologies we were using were unfamiliar to most of us. Lots of ugly code got written. Bad practices were rampant and often reinforced by our peers. The products were built and, for the most part, did what they were expected to do. But they were buggy, fragile, and nightmares to maintain.
Having a team of only senior level folks can have its disadvantages as well. Egos can come into play. There can be conflict as various folks attempt to establish themselves as the alpha-programmer. Differing and contentious opinions can cause conflict. Many teams in this situation often find themselves struggling to make progress, especially in the early stages of a project.
Another problem can arise when it comes to the pressures of schedules. If you find that given your current team structure, it’s apparent you’re going to miss significant milestones or deadlines, be leery of hiring more people to throw at the problem. There’s a quote, often called Brooks’ Law, from Fred Brooks’ book “The Mythical Man-Month” that says, “adding manpower to a late software project makes it later.” And it’s true. The best you can do in these situations is to figure out where the obstacles are for your teammates and do your best to eliminate them. Learn from the situation and when your team is readying for the next phase of development, adjust the team accordingly. Adding to your team late in the project will usually do little to get you to your goal faster, and will often just slow you down.
Leadership
Occasionally, you’ll find a software manager that promotes the idea of empowering team members to affect change. It’s very progressive and it sounds awesome. The thinking is that by empowering an entire group of people instead of a lone individual, the group as a whole will benefit from a large pool of good ideas. Not to mention, everybody likes to feel empowered, right? The folly, of course, is that every team is chock full of “experts” and they all have different ideas of the “right way” to tackle any given problem. Guess what? We’re not all experts. Not all ideas are good or even practical. And it’s usually only the loudest that get heard. That sounds harsh, but it’s true.
Software teams need leadership that can make swift, deliberate, and informed decisions with the least amount of friction possible. This is true regardless if you’re part of a modern, self-organized team or one with a formally defined structure. The only way to avoid group-think paralysis is by having a small number of informed and respected individuals whose opinions and ideas the team trusts empowered with the authority to make decisions that affect the entire project and team.
The choice of leadership sets the tone for the entire team. A calm, good-natured leader can have a calming effect on the entire team, even during times of crisis. A loud, overly-aggressive leader can make the team feel anxious and panicky in the best of times. When choosing your technical leadership, remember that the demeanor of the technical leader can be just as important as their technical skillset.
Workflow
A clearly defined and communicated workflow is one of the most important things a software development team can have. It sets expectations and lays out the lifecycle for units of work. This should communicate, at a minimum, the following key points:
- The developer’s next step once they’ve completed a code change.
- How changes flow through the source code repository. If you’re using branches and/or forks, what’s the structure and how/when does code move from one branch or fork to another?
- How code reviews are initiated, as well as the responsibilities of both the original code writer and the code reviewer.
- When/how testers get their hands on changes.
- How builds work and how to tell if a given change has made it into a build.
- How progress tracked.
- How product versioning works.
- When a given unit of work is considered done. This may seem obvious, but many teams struggle with this concept. Is it done when the developer considers it finished? Or do testers get the final say? Do other stakeholders need to chime in? Is there a difference between “complete” and “accepted”?
Once you’ve decided on your development workflow, capture it in your tooling. Many issue tracking systems make this simple. Some systems, like Jira, have explicit workflow support baked right in. Other systems, like Rally, often require a little bit more work forcing you to explicitly define tasks for each phase in an issue’s lifecycle. However you decide to do it, whatever tool you use, make sure it’s easy for everybody, including testers, business analysts, project managers, etc., to see it and use it. It should accurately reflect the state of the work at any given time. The importance of this shouldn’t can’t be overstated. The bigger the team, the more important it is.
Many teams are in a perpetual state of flux. They leave their workflow a bit open-ended, allowing it to evolve in hopes that it’ll settle into a natural rhythm that feels right for the team. For teams of more than a few people, this kind of approach usually feels chaotic.
Define your workflow, declare it so, capture it in your tools, and then stand back and watch. Make notes of what does and doesn’t work along the way. Allow your workflow wishlist to build and institute changes only at logical breaking points that carry the least amount of risk to your team’s work (e.g., after a product release). Making small changes frequently will fatigue your teammates and will often do little to alleviate the confusion.
Remember that communication is the litmus test for your workflow. If your team’s communication frequently breaks down, that’s usually a sign that your workflow has flaws.
The Religion of Methodology
Every software methodology has had its day in the sun. Even the waterfall model, which is often looked upon with disdain by folks today, was once all the rage. Methodologies, like many things in the world of computing, are akin to religions. For some developers, Scrum is the only true way to do software development. For others, it might be Spiral. But we oftentimes forget that a successful implementation of a software methodology is dependent largely upon the organization in which it’s implemented.
There is no one-size-fits-all software methodology. Just because you’ve had a long love-affair with Scrum or Lean, doesn’t make it appropriate for your current team. There are usually forces external from your software team at play. And these forces can significantly impact your team’s ability to implement certain types of methodologies effectively. Believe it or not, there are some contexts in which Waterfall is still the most appropriate way to do things. *gasp*
Shoehorning the wrong methodology into your team results in a lot of wasted energy and can often be an exercise in frustration. If your team tries to implement a given methodology and realizes it’s not working, perhaps you should give it up and find something else. If you can’t find a methodology that’s a good fit for you team, don’t be afraid to make something up. The methodology you use doesn’t need a fancy name, an O’Reilly book written about it, or a yearly conference dedicated to it. The most important thing is that you’re putting into practice something that allows your team to work effectively, work comfortably, and maintain good relationships with other parts of the organization. That’s it. If you feel like your team has succeeded in doing that, congratulations! It ain’t broke.
On a related note, if you try to implement a particular methodology that doesn’t work, you probably shouldn’t continue to identify your team with that particular methodology. For example, if you try Scrum, and the only thing you manage to adopt is the concept of a sprint, guess what? You’re not really a Scrum team. Continuing to identify yourself as Scrum can result in a team identify crisis of sorts. Meetings can quickly devolve into what your team does that is or isn’t Scrum. Team members may start championing for practices that might never be practical for your team. If you know you can never truly do Scrum, stop identifying yourself as such and you’ll cut down on the noise.
Workspace
The working environment significantly impacts team member productivity, as well as their sense of satisfaction. A software teams needs accommodations for both private, focused thinking, as well as collaboration. Don’t mistakenly assume that those needs are in equal proportion and certainly don’t get the proportion wrong. And, of course, the worst mistake you can make is sacrificing one for the sake of the other.
Software teams are composed of thought workers. And while it’s true that there are frequent occasions where collaboration is required, I expect you’ll find most your team members’ time is spent alone and focused. As such, you should ensure your environment is conducive for that sort of activity. The more distractions there are, the more frustrated your thought workers will be.
In the early part of the 2000’s, collaborative, or open-concept, workspaces for software teams became extremely fashionable. Technology companies started paying attention to what was happening in Silicon Valley and felt the need to emulate it. Ignoring the fact that most Silicon Valley startups were lucky to even have office space in the first place, often finding themselves in lofts or shared spaces with other companies, the rest of the world’s technology companies charged forward and spent buckets of money renovating their workspaces. Down came the office walls (ironically, managers’ offices usually remained in-tact). Away went the cube walls. In came the foosball tables and beanbag chairs. Conference rooms were rebuilt with glass walls in an effort to maintain the open-space concept. Developers suddenly found themselves face-to-face with their neighbors. What do you suppose happened next? The sphere of distraction took hold. Things slowed down. The quality of work suffered. And people started working from home. They had to. It was the only way to get anything done.
Unfortunately, many companies haven’t gotten the memo that the Great Collaborative Workspace Experiment was a failure and continue to push forward in the name of innovation. To those of you in positions of influence in these companies, I beg of you, be considerate to your thought workers and their needs. If you’ve created a working environment that you couldn’t do your taxes in, you should probably rethink your workspace.
The extreme opposite of open-concept workspaces are cubicle farms and private office spaces. There are plenty of faux pas to be made there as well. The first being shared space. If you’ve ever had to share a cube or office with someone, you’ll know that while it’s significantly better than an open concept workspace, it’s far from ideal. Distractions abound. Another problem is giving your developers a space that’s too small. No one wants to work in a closet. And if you’re an embedded developer, chances are your space will be cluttered with devices and various pieces of hardwaremabobs. These take room too.
Bottom line, your thought workers need privacy and adequate space to work comfortably. If they’re not comfortable, they won’t be doing their best. If you find that your hands are tied and that you simply can’t accommodate the needs of your developers, you may have to think a bit more creatively (hint: telecommuting).
Mentoring
I once worked for a company that hired summer interns, threw them in the deep end, and then expected great things. There was little orientation when it came to the codebase or tooling. There was certainly no mentoring. And worse, the interns were usually given the responsibility of implementing complex components that were critical to the success of the company’s flagship application. For all intents and purposes, the interns were treated like any other mid-to-senior level software engineer. How do you suppose that worked out?
Even if your team doesn’t hire interns, chances are you’ve probably got a few green software engineers on your team. It’s important that you recognize that not all software developers are equal. Those job titles (Software Engineer I, Software Engineer II, Senior Software Engineer, etc.) exist for a reason. Expect to see the same quality of work and pragmatism in your entire engineering staff regardless of experience level and you’ll be sorely disappointed with the results.
It’s critical to invest in your junior level software developers. Provide them with mentoring opportunities. It’s your team’s responsibility to teach them good habits and break them from the bad ones before they take root. Foster an environment of learning. We don’t graduate college knowing everything we need to know. I learned more in my first year on the job than I learned during my entire time in college. I’m sure for many of you, it was the same way.
Don’t forget what it was like being fresh out of school and eager to prove yourself. Remember the feelings you had when you first realized how much you didn’t know. And with that in mind, cultivate the environment that you wish you’d had when you first entered the workforce.
Junior developers won’t be junior forever. A lot of their personal growth hinges on the experiences your team provides. So make it count.
R&D
Many software development teams find themselves belonging to a larger organization-within-the-organization called R&D, or Research and Development. And unfortunately, many of those teams spend all their time on the “D” and very little on the “R”. Product development is important. But so is the research part.
From a business perspective, everyone understands the importance of investing in research. It’s all about diversifying and staying relevant. No company wants to watch their product line creeping towards obsolescence.
However, something that’s not often recognized is that devoting resources to research can also have a huge benefit to your software team. Software developers are a naturally curious breed. We don’t typically enjoy getting stuck on the same project for long periods of time. New technology excites us. Being able to work on something new and different from time to time can break the monotony and improve the overall sense of job satisfaction.
The way you allocate resources for research depends largely on your team’s commitments and priorities. I’m not going to suggest you do something as extreme as Google’s 20% time. But at the very least, you could encourage pet projects among your team members. I once worked at a company that formed a three-man “New Initiatives” team that was intended to develop proof-of-concepts / prototypes thought up by creative minds within the company. The plan was to rotate developers in and out of that team periodically. It didn’t really work for us, but it wasn’t because it was a bad idea. It failed mostly because it essentially turned into a product team overnight.
Regardless of how you choose to do the “R” of R&D, ensure that it’s a priority and communicate the status of those efforts to the rest of the team periodically. And most importantly, make sure each of your team members is able to contribute to the research efforts at some point along the way.
Training and Continued Education
I mentioned earlier the importance of investing in your team’s junior developers. This holds true for everybody else too. The software industry is changing at a faster and faster rate. Sure, developers need to be kept up to date with the latest trends, tools, and best practices. But so do business analysts, project managers, marketing folks, and the guys in QA.
Everyone on the team has a responsibility to continue to learn and develop their skillset. But the team’s leaders have the added responsibility to encourage their teammates and open doors to potential learning opportunities. Conferences, on-site training, online courses, books, trade magazines, and user-groups are all excellent sources of information.
Advocate for training and continued education to be included in the team’s budget. If money is tight, and it sometimes can be, encourage your team members to take the initiative and learn something on their own that they can present to the rest of the group. Weekly or monthly lunch-n-learns can be a lot of fun. Not to mention, they’re cheap.
Tooling
Software development teams use a lot of tools. Some of them are for the entire team, some are specific to the role an individual plays on the team. In all cases, it’s important for team members to know their tools.
On another team, I watched a project manager who used the same issue tracking tool for years somehow continue to fumble their way through it meeting after meeting. Could it have been because the tool was hard to use? Maybe. Could it be because the project manager never bothered to really learn the tool? Possibly. The real eye opener in this case was realizing the rest of the team also lacked a basic proficiency with the tool. Most never even bothered to use it. “How were issues tracked effectively?”, you ask. The answer is not well.
If your team is using a tool that it tolerates instead of embraces, that’s probably a sign you need to either switch tools or invest in training.
I would like to say a few things about developer-specific tools and builds, mostly because the developer’s point-of-view is my point-of-view, but also because developer tooling is essentially up-stream tooling and has an implied impact on the entire team.
Let’s talk about builds first. There are essentially two types of builds – developer builds and build server builds. Ideally, these should work the same. There will already be plenty of head-scratching opportunities when things work in one place, but not the other. Why add to the complexity by introducing different mechanisms for performing builds?
Developers should find performing local builds as easy as pushing a button or issuing a simple command. That one button or command should automate the entire build process. This includes things that happen before the code compilation process, such as image and font generation, language translation support, resource file compilation, etc, as well as things that happen after code compilation, such as running automated tests and building installers. This provides a consistent way for your team to build the software and codifies any assumptions made about the built components. It also effectively documents the way the software is built.
In addition to easy software builds easy, developers must be able to debug without too much effort. Don’t let your build process or tooling get in the way. As soon as you require a developer to think about how to use his tools, you’ve derailed his train of thought. You’ve distracted him from the real problem he’s trying to solve and a distraction like this can be a huge source of productivity loss.
A good rule of thumb is that developers should always be able to build and debug software from their local development environment. If there’s a separate team responsible for managing tools and builds, they should also be contributing developers who are able to eat their own dog food. They need to feel the same pain as everyone else. If they don’t understand how developers use the tools, they’re in no position to manage them.
Software Architecture
There’s a trend among some software teams, especially young Agile teams, to forego high-level, architectural design in favor of test driven or “just in time” design. The misconception is that big-picture design is a waste of time since things will probably change anyway and that you can refactor architecture as you encounter bumps in the road. This is a tad naive. Agile does not preclude architecture. To quote someone much beloved by the Agile community, Uncle Bob Martin,
“There has been a feeling in the Agile community since about ’99 that architecture is irrelevant, we don’t need to do architecture, all we need to do is write a lots of tests and do lots of stories and do quick iterations and the code will assemble itself magically, and this has always been horse shit. I even think most of the original Agile proponents would agree that was a silliness.”
-Uncle Bob Martin, Coplien and Martin Debate TDD, CDD and Professionalism
Test driven design has its place, but that place is at the micro level. Function implementations, classes and interface designs, algorithms, etc. all benefit from test driven design. At the macro level, test driven design is a recipe for disaster. A software team needs a clear and unified vision of what they’re building. It’s critical to map out at a high-level the components of a system along with their contracts and responsibilities. This process is best performed by one, perhaps two, individuals and not the entire team.
If you fail to perform architectural due diligence early in your projects, you will most certainly regret it. Refactoring architecture is not a trivial task and it can introduce a significant amount of risk, usually at inopportune times in a project’s schedule. You’ll do your team well to give this some thought up front.
I debated including software architecture in this list because it’s a bit more techie than I wanted to get. But a poorly architected piece of software affects the entire team, not just the developers. While developers feel the effects of poor architecture when they dive into the code, the rest of the team feels poor architecture in the form of slow progress, higher incidence of regressions, developer lethargy, increased technical debt, scheduling problems, inability to plan ahead, scope creep, etc. Avoiding architecture early in the project will give you plenty of short term gains. But in the long term, it’s like driving with flat tires.
Documentation
Every team generates some amount of documentation. Requirements documents, software architecture and design documents, build and code quality metrics, team process specification, meeting notes, HOW-TO’s, the list goes on and on.
Whatever types of documentation your team does or does not produce, it’s important that your team knows what sort of documentation is available and how to find it. It’s also just as important to recognize what is lacking and should be created.
Types of documents that are often overlooked include things like project requirements (“What!?” Yes, you’d be surprised how often this doesn’t exist), getting started guides for new hires, and workflow/process documentation.
Take a hard look at the documentation you do have and then imagine yourself as someone new to the team. What would they need to know?
It’s not uncommon for teams to amass a large volume of documentation that’s scattered across a variety of locations. Keep your documentation organized and easy to access. Whether you use a Wiki or a shared network folder full of Word documents, it doesn’t matter so long as it’s easy to get to. Missing documentation, documentation that’s hard to find, or documentation that contradicts itself all work against you.
Conclusion
If you recognize dysfunction within your own software team, take comfort in knowing you’re in good company. We all experience it in one degree or another. The real challenge is in not becoming apathetic to the cause.
The number one goal for a software team is to develop reliable, maintainable software that meets or exceeds our customer’s expectations. Team dysfunction is an obstacle in obtaining that goal. The bigger the team, the more opportunity there is for dysfunction.
I encourage you to think about some of the points mentioned above. And I invite you to share some of your own stories and ideas for mitigating team dysfunction. Please leave your comments below.
-Shane