Tuesday, January 29, 2008

Developers are not the only fruit


My younger self would be horrified to hear me say this but, when you develop software you need people who are not software engineers.

Don’t get me wrong, coders/engineers/programmers are the most important people because these are the people who actually produce the thing. As such they are central to any development effort, you can’t write software with a collection of managers, analysts and testers any more than you can build a ship without shipwrights.

Coding is where the metal is bent, shaped and welded. When coding there is no room for ambiguity or fudge, the code doesn’t lie and it doesn’t take to ambiguity. Thus there is a lot of details in coding which aren’t obvious before you start. So your developers need to be able to call on Product Managers for more detail and, if you have any architects, you need them at the code face coding with your developers.

See Jim Coplien and Neil Harrison’s patterns Work Flows Inward and Architect Also Implements for more discussion of this topic.

One of the worst projects I ever worked on was completely top heavy with managers (programme, project, customer, delivery, team - you name it we had a manager for it), testers, analysis, admin staff - yes you needed a lot of admin staff when the project was over 120 people. But only about 12 actual coders. Crazy. The overhead was there because... well, the ISO-9000 procedures needed a lot of managers, and it gave the consultancy implementing the project a lot of bodies to bill for, and it gave the Government (the people who were ultimately paying for it all) a lot of reassurance that it would be done one time, to budget and spec. Except, it was late, the spec trimmed left, right and centre, and if there ever was a budget we blew it.

So you see I’m no fan of carry extra weight on a project.

But, you need people who don’t code.

I’ve written before about the role of the Product Manager and why product management is important, and why you need them. So lets say we accept that Product Manager are needed.

In an idea world we wouldn’t need software testers either. And I expect one day to work somewhere where there are no testers. I don’t believe this is an impossible dream. But with the current state of play we need software testers in most organizations.

One role I don’t think we need is Software Architect. On the whole I think most architects are mislabelled. They should be Product Managers, Business Analysts, Senior Software Designers/Engineers or even Project Managers. Sometimes this is because the organization misuses the title, sometimes its because they don’t understand it (after all, what is software architecture?) but most of the time its because of title inflation. You couldn’t give someone a pay rise so you gave them a nice title.

Although I don’t read Joel Spolsky very often I do agree with him about Astronaut Architects. The moment your architects stops coding its time to fire them. Once they stop coding they are a burden not an asset.

Project Managers are another group I don’t think we need. I accept teams need some leadership but Project Managers are completely the wrong people to. Project Managers are trained to analyse, dissect, use whips (badly) and emit hot air. Their are not leaders by nature or training.

The Agile world is re-inventing project management as coaching but unfortunately the Agile world is pretty messed up about coaching. Coaches are useful when you they get it right and they are genuinely improving the team rather than acting as Project Managers with a different title.

But, and this is where my younger self will have a fit: you do need someone to oversee and manage the whole development process. The big problem is very very few people can do this job right.

By way of explanation to my younger self: done badly this role is worse than useless, done right it can really improve things. Unfortunately most people do it badly.

Its done badly for several reasons.

Firstly, most Software Managers have never been trained in the role - or management at all. They ‘make it up as they go along.’

Second, some of these managers don’t understand software development, they come from outside the domain. Such people try and make software development fit their existing model of management but it never will. Therefore they are constantly trying to make a square fit in a round hole.

Third, many of the managers who come from inside the domain, and know how to develop software can’t give up coding software. To be a good software manager you have to stop being a coder. You have to learn to look at the world differently, you have to realise you can’t intervene to fix problems, you can only arrange things so other people can fix them.

When I was younger I once worked for a bank. The code I was working on was terrible and I had no support from the organization. I really really wanted my manager, and his manager, to come and code with me. I thought their big mistake was to stop coding. I was wrong.

Their biggest mistake was not to stop coding but to stop communicating with those who did code. The big mistake was to put too much space between themselves and the building process. Had they involved themselves more they would have better understood the problems and would have been able to arrange things so I could fix them.

If you are managing a software team you should be talking to them every day. You should be improving their working conditions every day. You should be allowing work to flow to them.

When you move from coding to management you have to give up part of yourself, you have to trust the people you manage. It is their code now, not yours.

So, in conclusion: it is wrong to think you only need developers to create software. It is also wrong to think you can code and do other things - you need the separation. But it is also wrong to have too big a separation.

Make any of those mistakes and you just moved one more duck out of line.