Estimates got treated as gospel and planning was deemed extremely important, but nobody was willing to make the time for it to be done right.

I think web apps with shorter release cycles and fast deployments can really benefit from a process like this. You just have to be willing to try out simple, light weight tools and iterate on the process just like you iterate on the code.

Most shops tend to just use what they know rather than attempt to hack the process to meet their own needs. Kudos to the guys at UserVoice for going against the grain.

> Individuals and interactions over processes and tools

For my part, I'm completely done with Agile and the Agile community, there is very little of value there any more. Those who did add value have moved on, tired of and pushed out by those looking to make a quick buck consulting or managers looking to justify their poor management skills. It's quite sad.

I'm on a small team now, doing something fun and cool. The process is leaner, faster, way more productive and without all the BS overhead of trying to be "Agile".

I don't think it's quite that bad myself, but I think it depends which bits of "the community" you find yourself in. There are still smart folk who do actually pay attention to the manifesto and don't wander around applying an "agile" label to everything around them and carrying on as before ;-)

We just turned every other 'project management' app off but Trello (goals/tasks/bugs/feedback) & Beanstalk (repo);

* We view design as an ongoing and continual process - not a phase. Having a "software design phase" just wouldn't make sense for us since it basically covers the whole cycle from when the idea comes out to when it hits the streets.

* We generally pair on development where at all possible, and when we are pairing we don't find that we get a great deal of additional value out of code reviews on top.

* When we don't pair for whatever reason we do code reviews, but do them on a regular cadence rather than as a phase (say every Wednesday afternoon). A bunch of stuff fits in this kind of pattern. We are sure to talk to users every couple of weeks. We do some usability testing every couple of weeks. It's not related to how stories flow so it's not on the board.

I know that my examples are based on system software development, but I have no reason to believe they cannot be applied also for web development.

Anything that's going to be secret sauce is going to need a lot of prototyping and experimentation, and might not even yield usable results. Oftentimes, it isn't even possible to make an estimate as to the complexity of what it is you are building -- all you can do is timebox it.

Mixing that in with normal day-to-day development feels, to my brain, like mixing concerns.

Once those prototypes and experiments start to produce tangible benefits, then incorporating them just becomes another story in the sprint.

I think that "incremental" innovations are very important for companies and sometimes a "incremental" innovation can end up being a "radical" innovation.

That doesn't match my experience - and I'm not sure I understand the logic. Why would these algorithms be impossible to development during an incremental process?

From my perspective having a continual software design process, rather than a phased approach, gives me more opportunities for innovative design - not less. We end up doing more software design work that way, since we have the opportunity to add "secret sauce" all of the time - rather than just in some one-off section.

I find I get more innovation out of teams that develop incrementally. That is, to quite a large extent, why we develop that way.

I assume you've tried Pivotal Tracker at some point too? I'd be interested to know what you like/dislike about your new workflow compared to Pivotal.

However my main gripe with Pivotal was always two things: 1) Strict adherence to a Scrum methodology (forcing me to use a certain set of columns, etc) and 2) The UI. It frankly drove me nuts. The small text fields, small text areas and lack of wiki-style formatting. Anything longer than a few words was impossible to grok.

We flirted with trying it out about the time we adopted Trello but enough people had bad past experiences, like mine, that we never gave it a try.

Thanks for replying!

In addition, how do you track velocity?

From the description it doesn't look like they're running in sprints, but deploying as stories get completed, so cycle time would be the thing they would track. No idea if Trello can do this out of the box. They also explicitly said that they're not doing project estimates - so they wouldn't need to track velocity in any case.

If it were me I might be interested in visualising the flow somewhere - more of a tool to help spot problems as to make estimates any more accurate - but I'm a fan of just doing that with a whiteboard as part of running the project.

Perhaps you mean how many people in total? I've seen lightweight systems like this scale out to 150+ developers. (Insert discussion here about Agile Program Management)

You really don't need a lot. Most enterprise project management systems are over-engineered by a couple orders of magnitude. (And they hurt much more than they help)

Usually it helps if you stay with fixed iterations instead of Kanban and do program planning all at once (Yes, 150 people all in the same room working out the next couple of sprints. Fun to watch and participate in!)

Other kinds of tooling issues become bigger in a large group, like how to integrate separate physical pieces, how many branches to have in your CM tool, how technical drawings will be cataloged, etc. But none of it is super hard, and lots of other folks have already gone down this road. The important thing is not to over-constrain your development environment just to make yourself feel safer.

The biggest obstacle to running a larger group like this is dealing with the parts that are counter-intuitive. It just seems natural with 100 folks that you'd have a detailed charge-code breakdown, for instance. Or that 100% capacity is something you want to achieve. There are a few more gotchas like this, but if people are open-minded, it can mean a huge increase both in performance and happiness.

The industry is moving towards plain kanban without the sprints. I like this idea with smaller groups, especially groups with lots of support work on existing code, but for larger projects creating something new, using sprints, having a fixed point of coordination for a lot of different stuff, works much better.

Note I didn't say you couldn't go total kanban at large scale, just that there was a trade-off involved with each path.

I mostly don't work with organisations of that size, but the technique I hear people use to get around that coordination issue is to have events that happen on a regular cadence, but don't have them tied to delivery in the same way sprints are.

So you have stories flowing but also schedule fortnightly planning meetings (or whatever) on top.

This fits with my experiences on smaller kanban style projects where we have regular cadences for non-flow things like usability testing, backlog grooming, etc.

[Edit: Which is what the OP folk are doing "The ideas on this board are reviewed once a week during our Inbox Review meeting"]

The interesting thing here is that the same problem arises at a larger scale as it does on at the team level: architecture. Not how pieces of the software is put together, but the architecture of the existing and future business environment.

So, for example, a large project might include hardware, software, embedded software, offshore testing, multiple vendors, fixed demo dates, fixed regulatory dates, and so on. You can certainly manage all these cadences separately, but for most folks sprints of a week or two in length give you fixed reference points to then optimize around.

Of course, it's easy to think up a scenario with a pretty simple business environment, say the next version of Google+. In that case, Kanban probably would work a lot better. Do the simplest thing you can, but no simpler.

I find that business architectural issues can be difficult for the teams to grasp. The Scrum guys have preached sliding stories under the door to the team for so long, there's an entrenched mindset that things like that aren't the team's problem. (Not everybody, but many folks think this)

Counter-intuitively, by fixing the planning and delivery times, you can simply the cadence discussion. It all has to fit somehow in this list of dates. It also makes it easier to simplify or add complexity to the business architecture without having to worry so much about impacts.

You might think: why not just go kanban with one fixed planning cadence? Kind of the best of both worlds. Technically this would be awesome. You run into some communication and messaging problems though. If your message is "It's kind of like Scrum only bigger" then you can send out for Scrum training and explain how it all works. People will take the training and then ask questions like "who is the product owner" which aren't perfect but at least they're in the right direction. But if all you have is a bunch of boards and hand-waving, it's very difficult to get everybody to have some kind of mental model on what's happening. It's just too much (unless you already have established some previous Agile practice). Folks have to have some sort of working mental model of how things are supposed to work. Going from something like RUP to full-bore program-level Kanban in a complex business environment is just too far. My opinion only, for what it's worth. Perhaps in the next few years we'll see people tackle this. One of the things we don't talk about enough is that we struggle quite a bit with teams doing Agile, much less programs. It's simple concepts, but application is a bear.

Curiously what I've been hearing from folk in this situation is that it's the separation of the cadences that they see as an advantage. With the large teams they were finding there was just "too much stuff" happening at a single point and it became impossible to manage. Moving to a kanban style, and moving the various end-of-sprint rituals and spreading them out into their own cadences helped enormously - but I would imagine that it would be very context dependent.

The other thing I find useful about the kanban style is when the cadences of some work doesn't match up well with the sprint cadence. For example the more generative UX work (user interviews, etc.) doesn't always naturally fit into that 1/2 week cycle. Sprint-ahead/-behind is an abomination in my eyes - so having separate cadences helps a lot there.

Going from something like RUP to full-bore program-level Kanban in a complex business environment is just too far

Heh. With RUP you can probably find enough things in method composer to evolve it into something kanban-ish ;-)

Personally I'm not a fan of the big-bang rollout of new processes. Never seems to work well. And having folks who can actually understand "It's kind of like Scrum only bigger" means you've already got folk with a fairly sophisticated knowledge (assuming that they actually understand Scrum that is...)

One of the things we don't talk about enough is that we struggle quite a bit with teams doing Agile, much less programs. It's simple concepts, but application is a bear.

Seems that it's pretty much all I talk about - probably why I keep getting sucked into coaching roles away from more fun thing :-)

The cool part about Agile is that it's not Scrum, ie, there is no certification authority or magic recipe. It's just a bunch of schmucks sharing what worked or didn't work.

I shy away from answers like "it's all Scrum" or "It's all Kanban" for just that reason -- too facile. In a 15-project program, you usually have several groups going strictly Kanban. Enterprise architecture is one (if you have that) and program release is another. These things don't have to sync up, they just need to be ready. So even my advice about using sprints isn't absolute. Advice in this area is a lot of generalizations. You get what you pay for!

If you have any experience with these larger structures, I'd love to hear about it. I'm always collecting tidbits like that. Thanks for the chat!

Likewise and ditto. As I said I would imagine that it would be very context dependent.

This is very interesting. We also use Google Docs for stories/specs and Trello for prioritization and roadmap. However, we maintain a separate bug tracker (Open Atrium) for our enterprise clients to report bugs. We'd also love to have a single prioritization list. Wish there was an ability to create a Trello list sync'd with RSS Feed from Open Atrium.

> It’s the product development version of the Hunger Games

:)

Ticket - a message a customer sends in about an issue or question.

Bug - an issue (that can come from any source) which we track and resolve.

We just Trello for bugs (which may be reported in tickets) but we use UserVoice Helpdesk (http://www.uservoice.com/helpdesk) for tickets. Helpdesk is designed around customer communication, Trello is designed around development.

Hope that clarifies! :)

One point I would make as a brand new start up - I'm not sure how willing/able I'd be to handle the uncertainty of not forcing my dev team to tell me the number of hours a fix/feature will take to implement. I say this because sometimes our road map will include a feature request that upon receiving an estimate for I then decide not to proceed with as the cost/benefit ratio doesn't add up. When this happens I de prioritise the task and put it in a backlog. Having said that, we're at the stage where every second counts; perhaps when we've got a steady stream of income I'll be less picky.

Many thanks for a thought-provoking, and wonderfully complete, post.

I didn't see where a star rating equated to effort. I'll look again.

Thanks!

If you try this sort of thing yourself what you'll often find is that you can get the cycle time of stories (the time from the story going on the board, to the time it gets deployed) to be fairly constant. It can take some work to get top this point, but it's possible and useful in almost all circumstances.

Once you get to this point you can pretty confidently predict that stories are 1week/1month/2months away from getting deployed - and you can use this as the basis you base customer estimates on.

Arlo Belshee coined this style and called it "Disney Line" planning (y'know - like the one-hour wait from this point thing). Arlo has a great video on this sort of approach to planning http://www.youtube.com/watch?v=6t4bZtnnQJA that might be worth a watch.

Is there anything like Trello that can be run in-house?

I actually think if you're trying to get people to buy in to a methodology such as this, it's easier to start with a physical board and progress to something like Trello. One step at a time.

If I have more than one up-vote to give... this would get it. I find people really underestimate the value of a physical board setup. I agree that it's a vastly better starting point than online tools for most teams if they're in the same location.

I confidently predict that if Pivotal is not helping you now, Jira will not help you in the future. What needs to change is how the team works, or maybe management's expectations of the team. Not the tool.

a) stories being too big

b) stories having to much variation in size

c) doing time based estimation rather that relative estimates or story counting.

Predicting the future is hard.

But what about a feature request from a customer that you don't plan on implementing in the next 3 quarters? Do you just delete those cards, or move them to some other backlog.

Once you take a card you “put your face on it” (assign it to yourself). Dev etiquette is that you should never have your face on more than 2 cards at a time: 1 major project and 1 minor.