Decisions, Decisions...

Had a bit of fun with the team today, playing this game from TastyCupCakes.org.

A group of celebrities are trapped on a sinking ship. The team can only rescue one at a time. They must decide, what order to rescue them in.

It's good fun, and it touches on a number of interesting points about how we prioritise. This applies, not just to sprint planning, but also how we work as individuals.

The team quickly realised that they needed to identify the "must haves", and soon after that, the lowest priorities. The most time was probably spent on the in-between.

In practice, we can get started rescuing the top few, and make a decision later on the middle few once we know a bit more. If that goes well, we might even save a "z lister".

Justin Bieber didn't make it I'm afraid.

Tell me the Problem, Not the Fix

An interesting take from Rick Austin, on providing teams with enough context, so that they can meet the business need. As opposed to Product Owners defining solutions...

"When we create new products or add new capabilities to an existing product, we typically do so in an effort to solve a problem. These product updates may often happen by delivery teams being told to go build these features without a clear understanding of the problem being solved. "

More here: Do you know where you are headed?

Thanks to Kelly Waters for flagging that one

The Myth of Certainty

I've posted before about "uncertainty" as an indicator for estimation. I have also been reading some blogs & tweets on how estimation is pretty useless. Check out #NoEstimates or have a look at Vasco Duarte's blog.

I can certainly sympathise with the views being discussed. People have an overwhelming desire for certainty. It makes us comfortable. But, what does certainty look like? Some of my colleagues, past and present find that only dates and numbers provide that comfort, in as much as they insist on seeing the number of story points for a project up front.

This is pretty useless in reality, as we are pinning our hopes on something that is not particularly reliable. The first thing Ken Schwaber said at my Scrum training back in 2005, was that "estimates are always wrong".
"So", ask the #NoEstimates tweets, "why do we still do estimation?"

If it is certainty that we crave, we should be looking for predictability. Team velocity was one way the Agile community tried to increase predictability, but that wraps an artificial concept (i.e. points)around measuring output. There seems to be some momentum around doing away with story points altogether.

If a team is focused on delivering value, they should always be giving the customer what they need nice and early. Agreeing the MVP or MMF is a good start. Then the team should be breaking down their stories into chunks that they know they can deliver within a predictable time-frame e.g. 1 day. Because they do this, they know that they can deliver, say, 10 stories a week. It is proven, and predictable.

I am going to watch this debate with interest.

Test Early and Test Earlier...

Where does the testing process start? It is quite a change in mindset to get away from the concept of testers as "Defect Police. I've had a good few battles and it can take a long time to alter behaviours.

My mind is drawn to a conversation I had about writing test plans before development started. The QA in question looked shocked. "If the developers know what's going to be tested, they'll just write code to pass those tests. How will I catch them out?" It was then my turn to look shocked.

To me the role of a testing is way broader than catching out software engineers. If anything, they are the guardians of the product. Nothing should adversely affect the product, be that defects, or badly thought out requirements.

We do story analysis "just in time", rather than spending hours in painful and (I would argue) less than effective planning sessions. However, that does not mean that we should rush things.

It is important that we QA the story itself, before we start thinking about how to implement it. Testers have the expertise in finding the cracks. Looking for the edge cases, asking "What if..?". They should be looking at stories and asking the difficult questions, so that Product Owners are confident that their needs are being met, that there is a likely ROI, and developers are getting requirements with clear aims and without ambiguities.

London Lean Kanban Day March 2013 Part 5

I've already spoken to folks about the way Spotify work with teams, but Anders Iversson added a little depth to what I'd already read.
You can download a similar presentation from here.

This then segued in to the keynote speech from David J Anderson.
The talk generally centred around the ideas in this earlier presentation
Not that I like the idea of measuring how Kanban we are, but you could get team members to identify where we are in our journey, and compare notes to see where we need to strengthen.

London Lean Kanban Day March 2013 Part 4

Dan Brown is a bit of a NASA fan, but his talk did help focus on why incremental delivery is a good idea. Imagine if the moon landing was a "Big Bang" release?. Frightening. We need to try things, and we need to fail. Failure at a small level is manageable. Failure at the end of a large project could be catastrophic. This ties in a bit with Dave Snowdens talk on complexity. We need tasks that are "safe to fail", to ensure that we are going in the right direction.

Dan also showed us some maths.

Now, I'm going to take the formulae from here, as everybody knows I'm rubbish at maths.

The team is averaging delivery of about 12 stories per week.

We don't limit WIP seriously enough.

Lead Time is the time between the initiation and delivery of a work item. Currently LT=10.7 days
Cycle Time is the time between two successive deliveries. Currently, CT =0.4 days
Throughput is the rate at which items are passing through the system. Currently TP=2.4 stories a day
WIP – Work in progress; the number of work items in the system. Work that has been started, but not yet completed


And the question is how many stories do we have in progress at any one time?

WIP = Throughput x Average Lead Time
25.68 = 2.4 x 10.7

Therefore, as there is no cap on WIP we are working on 26.68 stories at any one time.

What happens to Lead Time, if I introduce a WIP limit of 4?

Average Lead Time = WIP x Average Cycle Time
Average Lead Time = 4 x 0.4

Therefore, Average Lead Time = 1.6 days

WTF!!!!?????

The formulas used on both solutions are equivalent:

WIP = Throughtput x Lead time
<=>
Lead Time = WIP x Cycle Time

This is an application of Little's Law, which I mentioned in a previous post. As I said, I'm crap at maths, so I'd love someone to have a go at this as well.

London Lean Kanban Day March 2013 Part 3

On a rather less Earth shaking note than Dave Snowden, Pawel Brodzinski has been working on how to use Kanban visualisation to manage programmes. It is a nice concept, and one I will be chatting about with my programme manager. It certainly resonated with me, as sometimes we don't really grasp the whole scope of how projects interact across the organisation.

London Lean Kanban Day March 2013 Part 2

As I settled in to prepare for more on WIP limits and how wonderful we all are because we call ourselves Lean Kanban types, Dave Snowden woke me up. Now, I had heard of Cognitive Edge, but I didn't know what it was about. Frankly, I still don't. Dave is a chap who will happily walk into the Pentagon (Therefore think what he's like in front of a bunch of wimpy geeks) and tell them that they know nothing and their thinking is completely wrong. Throw away all your cards and stickies. They won't help you because, by the time you have your stickies on the board, you've probably made your mind up about requirements, and you're probably wrong. I'm really not qualified to rabbit on about what he said, so here's his blog post on the talk.

London Lean Kanban Day March 2013 Part 1

Well, it makes a change for me to get out and listen to the great and the good of the Lean thinking world. The first Lean Kanban day in London, provided me with a good opportunity. The line up was impressive, and I have to say I enjoyed the speakers and came away with some renewed vigour. I will endeavour to cover the main points I found interesting in these posts. I'm going to to a post per speaker, and then maybe do a round up. That said, I may well change my mind half way through! First up was Mike Burrows with a general take on getting started with Kanban. This was the start of the "Big Thing" that I would be taking away. Understanding queuing, and the affect this has on our throughput. Whilst I have paid "lip-service" to WIP Limits, I have to say, I have struggled to convince people of the effectiveness. Perhaps that's because I hadn't been convinced enough to stick firmly to them, and therefore haven't been effective in convincing others. Mike referred to Little's Law. Sorry it's a link to Wikipedia, but it's a place to start. Now, I haven't read up on this fully yet, but there seems to be some mathematical truth kicking about here. I've had WIP limits demonstrated by playing manufacturing games, but that isn't software development. However, there were some other arguments in store for me. Slides are here

Story sizing and the "Uncertainty Principle"

I often have conversations with Teams about sizing, and more and more, I come back to uncertainty being perhaps the most important gauge for sizing. We use Fibonacci numbers, and when asked by people, "How do I size a story?", I go something like this... 1 - is the smallest size, and a 1 story has the characteristics of being quick to implement as there isn’t much coding involved AND we are certain about what needs changing and the impact. 2 - may need more work, but it is still well understood and we are still certain about what needs changing and the impact. 3 - we are introducing some uncertainty. We pretty much know what to do, and we know it’ll take a few days, but there is a greater risk of unknown factors impacting the work. 5 - we have a fair idea what to do, but we are likely to uncover some problems and so we are not certain about completing within a few days. 8 - there is a lot of uncertainty. It may well be best to create a “spike” story to work out what needs doing, and break in to smaller stories when we know more. Yes, we can add in other factors, and as teams become more experienced, they bring in more subtlety, but this seems a good starting point.