Geek Philosophy

(This post is a continuation. Please read part 1 and part 2 first.)

A year or so had passed since I first started implementing the tracking system I’ve been discussing. However, in August I started work on a larger project that had to go through the company’s formal project tracking system. The project was took about 3 months (we’re just finishing it off), and we have done no releases since – that’s an eternity for a team like mine that’s used to shipping production code very two weeks. With the project coming to a close, I was looking forward to moving back to a process that encouraged more continuous flow.

I had first thought about Kanban after hearing a talk at the Agile Vancouver mini-conference, Lean Development for Lean Times. I’d also recently finished reading David Anderson’s excellent book Agile Management for Software Engineering, which emphasized the importance of short cycle times and delivering value to the customer quickly. After corresponding with David about this book, I mentioned Kanban to him and was delighted to find that he was working on book on the subject. I was able to obtain a draft of this book to review, and the ideas presented within further cemented my ambition to try Kanban with the team.

Another thing worth noting is that my team had grown, from 5 engineers under me to 8. We’d also taken on maintenance duties for yet another product, bringing the total up to 4. This increase in responsibility caused me to spend more time thinking about our development process, since more was at stake.

Limiting Work In Progress

I’ve mentioned before that the only real difference between what I was doing before and Kanban is the presence of formal Work In Progress (WIP limits). What makes WIP limits so important? As it turns out, this minor change can have major effect on the effectiveness and responsiveness of the team.

Little’s Law, a finding of queuing theory, shows that the cycle time of an item in a system is directly proportional to the number of items in the system. Thus, to reduce cycle time, it is necessary to reduce WIP. This is intuitively obvious – I can finish reading a book faster if I read it all the way through than if I read some, then read the paper, then read another book, then come back to the first book – but by limiting WIP both stakeholders and developers are made explicitly aware of the implications of trying to do too much at once.

The benefits of limiting the size of the backlog have already been discussed in part, but by having a hard cap on the size of the backlog, stakeholders can clearly see that by choosing a certain work item other work items are not being addressed. This leads to increased collaboration between different business units, and forces everyone to “see the whole” (to use lean terminology) – individuals no longer argue for local optimizations (i.e. pushing their agenda through to the detriment of others), but instead come to an agreement about what would be best for the business as a whole.

Limiting the size of the backlog also helps stabilize lead time. Expediting items (“drop what you’re doing and work on this!”) is not conducive to flow, since it almost always involves context switching. Being able to provide stakeholders with the average cycle time of items is of great benefit – it can help make scheduling decisions, decide on relative priorities, and give visibility to other departments.

Limiting WIP in the “active” states (analysis, design, testing, customer acceptance) can help identify and deal with bottlenecks. For instance, if test is a bottleneck, which will be visible when test starts to run up against its WIP limit, we can use techniques from the Five Focusing Steps to alleviate the problem, such as ensuring that test is never idle by having a buffer of work always ready for them. Since the rate of flow through your system is limited by the bottleneck, improving the performance of the bottleneck will directly lead to increased throughput and reduced average cycle time.

I don’t yet have any numbers or metrics for my team showing improvement – I only implemented Kanban a week ago. As soon as I have a large enough set of data draw some conclusions from, I will post my findings.

Any questions so far? I’m understanding all of this, but I haven’t had a chance to explain it to anyone, and I always find that explaining / teaching really helps me solidify my knowledge.

(This post is a continuation. Please read part 1 first.)

In my last post I briefly discussed the “kanban-lite” system that I had been using over the past year or so. I adopted this system for several reasons:

• As a team working on multiple projects, there were often multiple stakeholders with different goals who wanted some of my team’s time

Each of these stakeholders had valid reasons for their selections of issues that we should tackle. By getting together and having a weekly meeting to see what was really pressing, I could ensure that I was working towards what was best for the company, not just trying to placate whoever was yelling the loudest. In lean terms, this is known as “seeing the whole”.

• My team was known as the “Fast Track” team, since our primary mandate was to respond rapidly to customer reported defects and small features that could close deals

Project teams at Sophos often undertake projects that are anywhere between 3 and 24 months in length. My releases are more on the order of two weeks. Since I’m releasing software so frequently, I wanted a system that encouraged flow and ensured that we were still delivering high-quality releases. Having a visible board where each item’s state was clearly displayed made it very obvious when we should finish and polish in-progress stories instead of starting new items.

• I’d been reading quite a bit about Lean Software Development, which encourages flow and the elimination of waste (another term for non-value added activities)

A key technique of Lean is value stream mapping – identifying where and when value is added to your product, and pointing out areas that can be removed since they don’t add value. The visual card board is an easy way to show at least part of the value stream (the value stream really begins when a customer files a defect or a product manager thinks of a new feature), and I had thought about (but not gotten around to) recording the amount of time issues spend in the various states.

So, you may ask, why would I switch to Kanban if my system was already working? I’ll discuss that in the next post.

Last week at work I transitioned my team to a Kanban system (warning, pdf). I’ve been thinking of doing this for a while, and in fact a lot of the process that I had already put in place was very similar to Kanban, but a couple of recent developments made the time seem right to take the plunge and fully implement the system.

First of all, what is Kanban? Kanban is a term that comes from lean manufacturing that means “visual card”.  The basic idea of Kanban is to have a system that facilitates flow and pull-based scheduling by limited Work In Progress (WIP).  In software engineering, this manifests as a visual process control system such as a whiteboard / cork board with story cards on it, or an electronic version of such a board. This board shows what state these cards are in, such as “backlog”, “in development”, “in test”, “customer acceptance”, and “waiting to ship”. So far, this sounds like pretty basic fare for an agile team.

The real key with Kanban is the limited WIP. All states of the board should have a WIP limit. For active states such as development or test, this intuitively makes sense, as it’s generally agreed that multitasking is not optimal. For a state such as “waiting to ship”, limiting the work in progress ensures that code is always shipped to customers in a timely manner. For a state such as “backlog”, limiting WIP (as opposed to considering the backlog to be the entire collection of defects, suggestions, and new features for the product) makes it possible to stabilize average cycle time (the average time it takes for an item to move from “backlog” to “complete”) and greatly eases prioritization.

As I said before, this is very similar to a system that I had set up for the team over the past year or so.  I would informally limit work in progress (by gathering a list of relatively high priority items in the backlog, encouraging team members to finish off in-progress stories before starting new ones, and shipping frequently). I encouraged a “pull” system instead of a “command and control” style approach, where team members would have a clear idea of relative prioritization and thus be able to choose the next task to work on when they’ve finished their own. I had a card wall that showed the various states that items were in.

What lead me to create a system that was similar to Kanban before I had heard about Kanban? I’ll talk about that in my next post in the series.

This post is a slightly edited version of an email I sent today in response to the question “Do you know any publications on performance loss due to switching from project to project?” I thought I gave a good answer, and I’ve wanted to write more about software engineering recently, so here it is.

It really depends on what is meant by “performance loss” and what the nature of the “switching” is.

The worst case scenario (while still remaining plausible) would be switching back and forth between 2 (or more!) projects in some sort of phased SDLC – that is, doing all of the analysis for project 1, then project 2, followed by all of the design for project 1, then 2, then coding & 2 and testing 1 & 2 and so on until you’ve shipped both projects. This may seem absurd to some people, and would definitely be a major process failure for an independent team, but in some cases this might be necessary; say, if specialist resources like designers and testers are a constraint, or if there are external dependencies. The obvious problem here is the greatly increased lead time for both of the projects. If the projects are of roughly the same size, the lead time has at the very least doubled (and that’s discounting any cost at all for context switching). This is a “Bad Thing” (there should be ample references in the lean literature about the benefit of decreased lead time), even if the scenario is not as blatantly bad as the one I just laid out.

For “performance loss”, it really depends on what’s being measured. If you’re thinking about lead time (which is directly related to throughput¹, and thus revenue – in other words, you should always be thinking about lead time), switching (or expediting, a different manifestation of switching) will negatively affect performance. Tom Demarco argues in Slack that the real penalties for task / project switching are often hidden because of how people measure efficiency (number of hours worked / busy people as opposed to value delivered to customers, i.e. throughput).

As for actual loss due to context switching, which was possibly the original intent of the question, I think this may have been partially covered by Brooks in The Mythical Man Month, and according to a blog post by Jeff Atwood of Coding Horror this topic is discussed in Gerald Weinberg’s Quality Software Managemnt: Systems Thinking. I haven’t read this book, but the title is evocative of Peter Senge’s The Fifth Discipline: The Art and Practice of Learning Organization (which helped pioneer systems thinking). I haven’t had a chance to read that either (it’s on my to-do list!), but it’s pretty well regarded from what I can tell.

So, does anyone out there have some other references to exactly how switching from project to project can cause trouble? I’m looking for papers, books, studies and the like, not just restating of lean or agile principles, so you can’t just say “context switching is bad because is slows people down” – show me the study!

¹Agile Management for Software Engineering, David J. Anderson [2003]