I should have been at Tom Gilb’s annual gathering this week - GilbFest as some people call it. Instead I’m sitting at 36,000 feet somewhere over Newfoundland heading for Chicago. I don’t often get out of Europe but maybe my fame is spreading.
The theme at GilbFest this year was Principles - defined by my dictionary as “a fundamental truth or proposition that serves as the foundation for a system of belief or behavior or for a chain of reasoning.” In fact principles is something that has been on my mind a lot over the last year or so. My “What and Why of Agile” which I gave to BCS London a few months ago featured a few thoughts on principles and I started a blog entry on the subject but never finished it.
Now is the time to finish it. Had I been at GilbFest this is what the audience would have heard about, or rather, this blog entry is the first of three instalments on what I would have said.
Let me say I don’t care whether you call these Agile or not, since we can’t define just what Agile is I’m prepared to accept they may just be principles. That said, I will divide the list into “Software Development Principles” and “Agile Software Development Principles” because I think the first set are universally applicable to software development while the second set require you to at least accept the idea of Agile.
I’ll publish these as separate blog entries, there is a lot here. After that I’ll also return to Kelly’s Law which I penned some years ago - before blogs!
Software Development Principles:
Principle 1: Software Development exhibits Diseconomies of Scale
Many, if not most, of us have been brought up with the idea that if we “do” bigger things get cheaper. Buying 2 litres of milk is cheaper than buying 1 litre. Building 1,000,000 identical cars is cheaper than building 10 different cars 100,000 each.
In software this isn’t true. Bigger teams are more difficult to manage, more expensive and less productive per head than smaller teams. This effect is so pronounced that really large teams might be less productive in total than small teams.
For example: a team of five will, per head, be more productive than a team of 25. Still the team of 25 will achieve more in total than the team of 5. However a team of 50 might be less productive than a team of 25 in total.
And its not just teams. Large software releases are more expensive than lots of small releases. Producing software to satisfy 100 users is more expensive than producing software to satisfy 10, or 1.
The effect appears again and again. Its why Lean folk like to emphasis small batch sizes. Unfortunately post-industrial society has internalised the concept of mass production and economies of scale. You, me, everyone, needs to purge themselves of economies of scale thinking and embrace dis-economies of scale if you are going to be be successful in this world.
By the way, I suspect this applies to other industries, more than we realise, however, I am a software guy, I can talk with authority about software so that I’ll stick to there.
Principle 2: Quality is essential - quality makes all things possible
By quality I’m really thinking bugs, I want to see bug-free software. I definitely do not want to see gold plating, I have no time for reusable code (as I said in an earlier blog).
Philip Crossby said it best: “Quality is Free” - Neils Malotaux puts it more accurately if less dramatically “Quality is cheaper.” The basic message is the same: pay attention to quality, rid yourself of rework and it will work out better. I said more about this in my “How Much Quality can we afford?” presentation.
And when we get to Agile I’m quite clear: if you don’t build in quality I don’t see how you can get iterations to work and I have no hope you will ever truly achieve Agile.
I’ll finish here for now, I’ll continue with the Agile Principles in the next entry.
To finish I should say these principles are a work in progress. That doesn’t mean that I intend to change them when things get tough. Rather I mean a) there may be some more I haven’t get identified, b) there might be some even deeper underlying truth below some of these principles.
Boehm also uses diseconomies of scale in his COCOMO models. He uses SLOC as his measure.
ReplyDeleteQuoting, "consider the cost of producing two software components having 10,000 new SLOC. If the development cost is $20/SLOC, the cost of each will be [...] $200,000. But suppose that these two components need to be combined into a single product. In addition to the $200,000 it costs to develop each component, the project will incur additional design costs to make them interoperate, integration costs in order to deliver them, testing costs on the integrated system, and rework costs to fix any integration defects. The total number of SLOC is still 20,000, but the cost will be greater than 2*$200,000..."