In his Reinventing Business presentation at geecon last month, Bruce Eckel suggested that improving how companies operate requires experimentation. Since then, I’ve been wondering whether the absence of dedicated project management at Lunatech is a reproducible experiment, rather than a random exception to the norm in software development.
Hypothesis: transparency and better communication let development teams be self-managing.
Corollary: this is the start of a project management renaissance, as project management gets more interesting.
In summary, this article asks a question: can other software development organisations use a policy of transparency, varied Internet-based communication tools and the occasional programmer with social skills to remove the need to have a dedicated project manager for every development team?
On a small cargo ship, with a crew of five or six, there’s the skipper (captain), the first mate (would-be captain), a few deck-hands (able seamen) and the engineer. Unlike the skipper, who also has management functions, the engineer is a senior role purely because he is a specialist in running the ship’s ultimate back-end system: a giant diesel engine.
Ten years ago, I was a programmer - deck hand if you like - on development teams that typically had five or six members. There were similar roles, although 'would-be project manager' was not explicit. Senior to the programmers was the DBA. (I hesitate to say ‘Database Administrator’ because it suggests a more limited scope and status than what I experienced.)
The disappearance of the DBA
I haven’t been on a cargo ship for years, but I have noticed that during the last ten years, the DBAs went away. First, I worked on projects that didn’t have their own full-time DBA, but hired one in for a few days per year. Later, I realised that I hadn’t seen a DBA anywhere on a project for several years. What happened?
I don’t know what happened. It occurs to me that the way we use technology has changed somewhat, so that the database is a less central component of what we build, even though it’s always still there. Also, we switched to open-source databases that over the same ten years have introduced two important qualities: they Just Work and are easy to use. Meanwhile, I expect that the DBAs are somewhere else, working on that small subset of software that represents a hard database problem. Ships haven’t changed; software has.
The same thing has happened to the dedicated system administrators, as it happens: now you don’t necessarily need a full-time specialist to deploy software in production.
Team ÷ 2 - 1
Similar changes have affected project management at Lunatech: transparency, communication tools and agile software development.
After Clay Shirky invented the Internet and transparency (not really - I’m summarising), software developers realised that they could use better communication tools, more transparency and sharing the workload to make a third of project management redundant: information-sharing. Status reports and status meetings are gone.
Agile software development has changed too, in that more of our customers embrace rather than resist it. The change on the team is that simpler agile methods facilitated by one or more developers (e.g. a Scrum Master) make another ‘third’ of project management redundant: managing the process and assigning work. Now developers self-assign work.
The first big consequence of this is that there is less project management left: mostly stakeholder relationship management. Given the occasional developer with good social skills, this becomes one of the developer’s part-time responsibilities. The second change is that a lot of the above, including agile software development and simpler architectures, mean that we can solve the same problems with half the number of developers.
The end result is that we replace the old six-person team that includes a dedicated project manager, with a self-managing team of three developers that includes one who spends around 30% of the time on project management.
Part of what makes the development team smaller is that you kill off empire-building behaviour. There is a risk that a full-time project manager won’t have enough to do unless the project is in trouble or there are ten people to manage. Most project managers prefer the bigger team. However, if project management is someone’s part-time overhead function, they are likely motivated to do it as efficiently as possible, to have more time to produce deliverables.
At Lunatech, we don’t need to worry about where this leaves pure-play project managers, because we don’t hire them. I imagine that the good ones find big or difficult projects that need a project-management specialist, or that they diversify and broaden their roles to include business analysis or product management (e.g. Scrum product owner) roles.
The concluding question
I suspect that the accidental evolution of half-size self-managing development teams at Lunatech is something that might be reproduced elsewhere. Does this sound plausible? Is it obvious how you might go about this particular experiment, or would you need more detail on which specific practices can make this possible, such as exactly how you get rid of status meetings and customer status reports?