Yesterday I had the good fortune to have a chat with Woody Zuill over Skype. I was a really nice hour that just flew by, where we exchanged stories, ideas and had a few laughs.
Let me, as a side note, just say that I love that; just exchanging stories with people. I always learn new stuff and quite often stories from my own experience pop back into my memory. This is the best part of conferences - meeting, interacting with others. That’s what I miss the most, being a little sidestepped here in Indonesia. Praise the Lord for Skype and Twitter… ah well for the Internet too (although I’m not entirely sure He had any hands-on action on either of those inventions).
I just thought I summarize some of the things that I took away with me from our conversation.
It was never about the estimates, really
Tell me what you want, what you really really want! I wanna wanna wanna … an estimate?
Woody is the man behind #NoEstimates among other things. This has proven a highly controversial topic and I have from time to time refrained from talking about it… out of fear or at least tiredness of being bashed about this on Twitter and elsewhere.
A few things to note about #NoEstimates, only speaking from my own experience, that I think sometimes are missed:
I never heard about a customer, you know the guy who are paying our bills, thanking anyone for the awesome estimate. We (think we) need them to do our work, but that also mean that we can change that. Since it’s nothing that the customer wants we could do less of that and more of the things that the customer do wants.
Processes are always only best so far
Secondly, this goes for anything we do. The customer doesn’t want requirements, plans, standups, tests or even the code. The customer wants a problem solved. All the things we are doing to solve the problem for the customer is just “best so far” and we should change it as soon as we are creative enough to look outside our current understanding.
For some, me included, looking beyond our current ways and practices is very hard to do. That’s one of the reasons I always react with doubt and question introduced to new concepts. I’m a programmer so this list is a bit skewed, but this is my exact words or thoughts when I heard about some concepts that I now have come to appreciate:
- WHAT?! Two people at one keyboard. That’s just stupid! (Pair programming)
- Meeting every morning? Shouldn’t agile mean less meetings? (Standups)
- Writing the test before the code? That super hard to do, and it will produce a lot of test code that just testing the code (TDD)
- Dependency Injection is just for computer science researchers. I’m coding here - I don’t need that (brrr totally true I’m afraid)
- Open source?! But we need someone to take responsibility for the maintain of the code we’re using
- Mob programming?! This is just wrong - all of the team in the same room. With one keyboard!
- Chaos monkey - why on earth would you write a program that is causing problems. We need stable IT here. Who did that? (Netflix dudes, the biggest infrastructure/application on the planet)
I could go on and on but I’ve probably bored you already. You probably are much better to assimilate new ideas. Two things that I’m proud of myself for: I’m faster to go from “WHAT?!” to “ok, this might be something” nowadays - and I recognize many of the things that I say about new ideas, as things that people say to me when I introduce them.
I think that one of the biggest reasons that #noEstimates is so sensitive is that we’re trying to change one piece that we find so fundamental in how we do work that it’s a given. It’s how it’s done - how else could we [fill in your reason here].
Which brings me to 3
It was never about the estimates
No. It was about becoming better solving customer problems. It could have been #noRequirements #noCode #noDevelopers or what have you.
I often point to Toyota but since they came up with much of the things that we’re into right now… well I like them. One of the things that I really think they’ve got right is around their vision. It’s not about selling the most cars in the world, or being the most sought after employer. It’s about their process and goes something like this: “One piece continuous flow” meaning that from the time someone orders the car only value adding activities are done, just in time, with no inventory and no waiting times.
WHAT?! But that’s … oh, sorry. There I went again.
The thing is: it’s not about getting there - it’s about the journey. “One piece, continuous flow” is a compass pointing towards a wonderful state we probably never will reach. But we (ah, well Toyota’s 1.3 million employees) thinks that in trying to reach that goal we question the things that we are doing today and try to improve them for the better.
“But if we’re going to run this factory without inventory… that means that we need to have lorries coming into the factory gates with new material ever 15 seconds” - this happens in a Scania factory in Brazil. It took them a long time to get there, but they keep pushing for better. “But we’re going to deploy 50-100 times a day… that means that we totally need to change the way we build, test, deploy and verify our code” - this happens at many of the largest sites around the world today “But if we’re going to run without estimates … that means that we have to totally change how we plan, write requirements, interact with the development team, do budgets and even get certified/approved for governmental funding” - Yes. Still, this happens at many famous agile shops today. Pick up the Specification by example book, or check out Dan Norths talk on Embracing Uncertainty for example. Dan has worked for a big bank doing this… Bank of.. Bank of… yeah, that’s right Bank of America. But it only, only works there. Or does it?
Let’s leave the #noEstimates for now - it’s so scary to even talk about. I see pitch forks and torches out on the street…
((((DO SOMETHING!) SMALL) USEFUL) NOW!)
The above LISP (?) function is defined by Bob Bemer and I love it. Woody said something that got me thinking of that quote in our conversation: “I don’t care much about the order of the backlog or pay much attention in if we’re doing the most important thing”
WHAT?! But … now it’s getting old right? I’ve spent SO much time prioritising backlog, checking that we’re doing the most important thing. This provoked me!
But the point here according to me is that if you do something small often it doesn’t really matter that it’s the most important thing.
If you push things into production 2 times a year it’s of course super important that you’re doing the most important thing. But if you’re (like my last two gigs) push things 10-20 times a day, it’s not as important.
Doing things that small require a totally different way of working though, but consider how much we have changed in how we do software development since agile came into existence. TDD, Pair programming, meetings, source control, micro services, etc. etc. It’s “totally different way of working” (quote from Marcus Hammarberg, about 5 seconds ago).
Maybe we can become better if we try to let that spill over in how we test, how we write requirements, how we plan, how we est … (nearly did it again. Walk away Marcus. Don’t look them in the eyes and just walk. A. way.), even how we do business.
Go concrete early and adjust
On of the things that I’ve come to appreciate and try to live by in the last couple of years, when it comes to code, is to try to become concrete earlier.
This is in the relation to specifications and code where we make our shared understanding concrete by writing executable specifications and implementing the fully fledge feature in small increments so that we could try them and validate our learning earlier.
I like this very much and as I’ve have had this in my head for so long I have started to try to apply in other areas as well; how do we make an new organisational structure concrete earlier? Can we do simulations or dry runs, walking around and pretend that we’re in the new organisation and see how it would play out? Could we implement it in just a small part of the whole organisation?
I think there’s a great meta-learning here trying to become more concrete earlier. What should we change in the current way we are working in order to let our ideas become concrete earlier?
Finally under this heading I was reminded about a story from Avega, a consultancy I worked for a couple of years back. We sent a team (not including me) to a programming contest.
The task at hand was to write a little program that went around a playing field and collected “gold”. The team with the most gold after 3 hours won. There was a set HTTP API that you called in order to move your “robot” around the playing field. When you had 10 golds in your bag the robot needed to go back home and empty his bag before he could go out again. This was the point where you could update your program too.
Our team won. They crushed the others. But they didn’t won because they wrote the most nifty gold-finding algoritm. The won because they iterated fastest. Here’s what they did:
First they created the simplest thing that could possible move the guy around. It was one of the team members (Joakim Sundén I think) that used the WASD-keys to move the “robot”. So it was not even a robot.
As the first version was created Joakim started to go around collect gold, manually. In doing so he didn’t only just started to collect gold, he also fed back ideas to the programmers
- “Guys, the server is lagging. Maybe we should make sure we get a response before the next command. They are queued up now and I run over the edge”
- “You know what would be great? If our guy could remember where he was when he had to go back home. He could then automatically replay those steps to get back to the same location”
- and more like that
The team worked on the new features in two groups and the feature (either one) that was completed first was deployed at the next round.
Every feature was created in the simple possible manner first (remember this still wasn’t a robot), wasting no time on the playing field, where Joakim was collecting gold.
If my memory serves me right it was not until the last round that the robot did automatic picking of gold. By then the winners where already clear; Team Avega had won. Also the other 3 teams aimed to implement an automated robot:
- 1 didn’t finish at all - 0 gold
- the other two was significantly slower to produce their first version on the playing field. Since that was their first version it had some severe bugs that slowed them down significantly. Hence much less gold than our team.
Team Avega won because they iterated faster and deployed something valuable faster. They did not build what the specification (game instructions) said, but took the aim of the game and use that as their inspiration to become better than even they thought they could be.
And in the process inspired me.
Constraints are great to produce innovation
Real short on this; I’ve blogged about constraints before. But when you listen to the history of Woody you here constraints come into play again and again:
- “I couldn’t afford to study full time, so I had to learn on the job meaning that I always pushed myself a little bit more”
- “Learning to program I only had 1 hour to spend each day and I needed to build stuff for myself so I …”
- “My firm couldn’t make big investments so we had to take on jobs that showed that we could finance the purchase expensive equipment before we went to the bank”
That only strengthen my thesis; with constraints comes innovation. The worst thing that could happen to a project, team or company is to have to much resources. I’ve worked in placed with too much money - disaster follow both in morale, quality and customer focus.
If you are in such company - add artificial constraints. This is essentially what WIP limits and sprint lengths are. And when they feel comfortable - lower them. This will push you further and further towards a better future.
Maybe even better than you can imagine now. Imagine that. Imagine wonderful.
Woody often announces that he has time available to chat. Don’t miss that opportunity - who knows what kind of thoughts a chat with a very experience agile fellow like Woody Zuill might trigger in you.
Thanks a lot for the inspiration Woody - looking forward to the next time. DS