The Best Way To Understand Agile

I was chatting with Jason Tanner of Enthiosys the other day, and he asked me how I defined “Agile Product Management.” What follows is more or less my answer, formatted a bit for the purposes of the blog.

First, what is agile?

All my discussions of “agile” anything start from how I describe agile at the 30,000 foot level, with the following rules:

1. Figure out what you want to do
2. Figure out which of those things is the most important (“predicting the future” step)
3. Do the most important thing until it’s done (“done-done“)
4. Go back to #1

Each time through that set of steps is an iteration or sprint, and you’re likely to have a lot of sprints before you get to a release.

I contrast this explicitly with the comparable rules for “waterfall”:

1. Figure out what you want to do
2. Figure out how to those things in #1 most efficiently (predicting the future step)
3. Do those things in the most efficient order

It’s a subtle difference when shown this way, but the difference is critical, and for a number of reasons:

• Each time through the agile process (each iteration) you have the option to define a new set of “things to do” – this means you can respond to the market, or realize that your original idea was wrong, or realize that you can’t actually accomplish your original idea, or refine the idea or design to be better, whatever it might be.
• All management disciplines suffer from the fundamental problem that predicting the future is hard or impossible. No matter how much you know, your predictions are likely to be wrong, and to get wronger as your prediction horizon extends. The benefit of agile is that your horizon for predicting the future is very short – what’s the most important thing to do right now? – while it’s very long in the waterfall case – “what’s the optimal way to do all these things?” As a result, the damage to the project from incorrectly predicting the future is much more contained with agile – to a single sprint or at most a few sprints. In lean, it might be contained to an even smaller increment.
  • (Sidebar: Why is predicting the future so hard? Three reasons – one is that things change, so what seems important today might not be important tomorrow; second, you might need working code to understand what you really want; finally, development estimates are never over estimates – everything always takes more time than estimated, never less time.)
• Agile explicitly supports the Pareto Principle (or the 80/20 Rule), while waterfall doesn’t have a good way to do that. In other words, in any set of things I want to do, 20% of them will give me 80% of the benefit, and in agile I make sure to do that 20% first.

Applying these ideas to product management

Given this, how do I define agile product management? The steps for agile product management are no different than they are for other agile activities (like agile project management, what we usually mean when we say “agile”). Figure out all the things you want to do, prioritize the list, and do the most important thing on the list, then repeat.

What does that mean in practice? It means don’t spend time on unimportant things. Always know what’s most important (how I determine that is covered in another post), and focus on getting that to “done done” before doing something less important. And then when you’re ready to start again, redo the list, and reprioritize. As a product manager, I make a list of the features needed by the customer or the market, prioritize it, and write detailed requirements for the handful of the most important ones. Those I then socialize with the engineering team, both to get their rough estimates, as well as qualitative feedback on the feasibility of the features, open a conversation on what the feature really is, get ideas and innovations from engineering that might make the feature more valuable, and so on. That’s one iteration – then I do it all again for the next iteration.

Contrast this with a non-agile approach to product management. I’d start with the same list of features for the system, then write all the requirements, in enough detail for engineering to estimate them all. Then, after a lot of discussions, covering all the requirements (let’s hope we don’t miss an important one!) I got the (wrong) estimates from engineering, I’d then lay out the tasks in an optimal way to get all those requirements delivered, balancing a desire to get full functionality with a need to meet certain market milestones. Because I’m predicting the future, the resulting schedule is guaranteed to be wrong, and it will always be wrong in the undesirable direction.

But there’s some stuff missing!

At first glance my definition of agile may seem simplistic, and even missing something – where is all the stuff that people usually think of when they hear “agile”, like Scrum, sprints, retrospectives, velocity, and all that?

My point is that agile is not about specific tools or techniques (although agile methodologies may define or specific tools or techniques) – it’s a mindset about getting the most important stuff done-done as quickly as possible.

But let’s expand it a bit more and talk about where all that “agile stuff” is – or rather, where it arises from given my definition. For example, there’s a claim that agile teams keep getting faster, and they work more collaboratively and generally are superior to teams that aren’t using agile. But what’s really happening is that if you run a project based on my definition of agile – whose primary output is to get the most important things done first – you get lots of additional desirable side effects. You not only have a much better approach to the project management side, but you get to take advantage from a management perspective.

If people are always working on important stuff, then they are going to be more motivated to do a good job – this will improve quality over time. And if we’re getting insight – and corrective action – into how things are going on a very short timeframe, like a two week sprint, then we can apply changes directly when they are needed, and over time we can apply a LOT of changes. Simply doing a retrospective every sprint, rather than every project, gives you a lot more leverage, for three reasons:

1. The problems are still fresh so it’s easier to see solutions
2. The solutions you apply early on benefit the entire rest of the project
3. You end up getting to apply a lot more solutions.

If you only do a retrospective at the end of the project, as in waterfall, then you’re going to forget some of the problems you had, you’re not going to be able to fix all the problems you do remember because the change will be too big to do at one shot, and only the next project will benefit, not the one that just finished. All in all, a nearly complete waste.

In fact, you might want to think of my rule #3 as having a sub-step – “3.5. Take advantage of your new experience to improve your process.” That will happen every iteration, and you’ll get better a lot faster.

That’s the primary way you get better as an agile team, although there are a few other well-known practices you can apply that will immediately start to give you benefits, like team programming and other collaboration approaches, making sure that everyone on the team takes responsibility for getting to done-done (meaning that developers sometimes do a lot of testing, and testers contribute a lot to development), and always having a deliverable product.

What do you think?

Does this definition of agile resonate for you? Do you use another formulation that captures the key driving benefit of the agile approach? Do you have additional questions about agile and how all the benefits of agile arise from this simple set of rules? Let me know in the comments.

I’ll post again soon about agile, specifically answering why these rules result in getting better quality products to market faster. That’s another promise from the agile camp, and it’s interesting to think through how it happens. And we’ll touch on how to talk to management about agile. Management has historically liked the waterfall approach, because people it appears to be predictable – although the predictions are always wrong (way wrong). Agile seems suspect to management, on the other hand, because you can only say “we’ll get the most important stuff done first,” but you usually can’t commit to when that will happen.