Thursday, December 12, 2013

Stop Telling Stories

There are beautiful, simple ideas in today’s Agile development methods that work really well. And some that don’t. Like defining all of your requirements as User Stories.

I don’t like the name. Stories are what you tell children before putting them to bed, not valuable information that you use to build complex systems. I don’t like the format that most teams use to write stories. And I don’t like how they use them.

Sometimes you need Stories, Sometimes you need Requirements

One of the “rules” of Agile is that stories have to be small – small enough to fit on an index card or a sticky note. They are too short on purpose, because they are supposed to be place holders, reminders to have conversations with the customer when you are ready to work on them:

They're not requirements. They're not use cases. They're not even narratives. They're much simpler than that.

Stories are for planning. They're simple, one-or-two line descriptions of work the team should produce.

This isn't enough detail for the team to implement and release working software, nor is that the intent of stories. A story is a placeholder for a detailed discussion about requirements. Customers are responsible for having the requirements details available when the rest of the team needs them.
James Shore, The Art of Agile - Stories

According to Mike Cohn in his book Succeeding With Agile, making stories short forces the team to shift their focus from writing about features to talking about them. Teams want to do this because these discussions are more important than what what gets written down.

But this idea can be – and often is – taken too far. Sure, most people have learned that it’s not possible to write correct, complete, comprehensive requirements specs for everything upfront. But there are lots of times when it doesn't make sense to limit yourself to 1- or 2-line placeholders for something that you hope to fill in later.

Some requirements aren't high-level expressions of customer intent that can be fleshed out in a conversation and demoed to verify that you got it right. They are specs which need to be followed line by line, or rules or tolerances that constrain your design and implementation in ways that are important and necessary for you to understand as early as possible.

Some requirements, especially in technical or scientific domains, are fundamentally difficult to understand and expensive to get wrong. You want to get as much information as you can upfront, so developers – and customers – have the chance to study the problem and think things through, share ideas, ask questions and get answers, explore options and come up with experiments and scenarios. You want and need to write these things down and get things as straight as you can before you start trying to solve the wrong problem.

And there are other times when you've already had the conversation – you were granted a brief window with people who understood very well what they needed and why. You might not get the same chance with the same people again. So you better write it down while you still remember what they said.

Short summary stories to be detailed later or detailed requirements worked out early – different problems and different situations need different approaches.

The Connextra Template: As a {type of user} I want {something}…

Stories started off as simple, free-form descriptions of work that the development team needed to do, like “Warehouse Inventory Report”. But now, the Role-Feature-Reason template also known as the Connextra template/format (because somebody working there came up with it back in 2001) is the way that we are told we should all write user requirements.

as a {type of user}, I want to {do something}, so that {reason}

How did this happen? And why?

According to Mike Cohn (who helped to popularize this template is his books and courses) there are a few reasons that stories should be written this way:

Reason 1

Something significant and I'm tempted to say magical happens when requirements are put in the first person…

Reason 2

Having a structure to the stories actually helps the product owner prioritize. If the product backlog is a jumble of things like:
  • Fix exception handing
  • Let users make reservations
  • Users want to see photos
  • Show room size options
… and so on, the Product Owner has to work harder to understand what the feature is, who benefits from it, and what the value of it is.

Reason 3

I've heard an argument that writing stories with this template actually suppresses the information content of the story because there is so much boilerplate in the text. If you find that true, then correct it in how you present the story… [which is not a Reason to use this template, but a workaround if you do use it}.

Trying to fit every requirement into this template comes with its own set of problems:

User story format is awkward and heavy-handed. “As a ___, I want ___, so I can ____.” The concept is good – there should always be an explanation of “why” the task is desired, to make sure the end result fulfills the actual need. But the amount of verbal gymnastics I've seen people go through to try to make a simple and obvious requirement into a “User Story” proves that despite what Agile says, it’s not always the best way to go.
Talia Fukuroe, 6 Reasons Why Agile Doesn't Work

Lots of other people have seen similar problems:

Steve Ropa at VersionOne (“Why I Don’t Like User Story Templates”) has worked with teams that don’t understand or properly follow core Agile ideas and practices…

But where they excel is making sure every story is expressed in the format As a ____ I can____So That____. No matter what the story is, they find a way to shoe horn it into that template. And this is where things start to fall apart. The need to fit the story into the template becomes more important than the content of the actual story. …

The Template has become a formalized gate. My understanding of stories when I first learned about them was that they were to bring natural language back to the conversation around what the software is intended to do. How are we to move away from formal “shall lists” and requirements documents if we are just replacing them with Story Templates?

Gojko Adzic says that robotically following a standardized story template leads to stories that “are grammatically correct but completely false”:

Stories like this are fake, misleading, and only hurt. They don’t provide a context for a good discussion or prioritisation. They are nothing more than functional task breakdowns, wrapped into a different form to pass the scrutiny of someone who spent two days on some silly certification course, and to provide the organisation some fake comfort that they are now, in fact, agile.

Everything is about the Customer

We are also told that User Stories must be customer-centric, written from the customer’s point of view and always describe something that customers care about.

This argument is built around two central ideas:

  1. The development team must always be clearly delivering Customer Value. From the start of the project, the team is supposed to deliver working features that customers can see, touch, explore and respond to.
  2. The Customer/Product Owner has to be able to understand every requirement, which means that every requirement must be in their language and something that they care about. (This is another example of what’s wrong with the Customer/Product Owner idea in Agile development – that a single person can be responsible for defining everything that is done in a project.)

Focusing exclusively on delivering Customer Value leaves little room for non-functional requirements and constraints, which are critically important in building any real system, and de-emphasizes important design problems and technical work that developer teams need to do in order to deliver a high-quality system and to minimize technical risks.

It has led to lots of unnecessary confusion and disagreement over how developers should take care of non-functional requirements and constraints, developers hiding technical requirements inside customer stories, or trying to warp their own technical and architectural requirements into customer-style stories that end up not making sense to customers or to developers.

It is especially a problem with under-the-covers technical requirements like security, maintainability and supportability – cross-cutting concerns that don't make sense to a customer, but are fundamental constraints that apply across all of the work that the team does and how they do it, not just work done in one Sprint.

Like the idea of a using common template, putting the customer first in requirements was well-intentioned: a way to bring the development team and customers closer together, and a way to make sure that the people paying for the work actually got what they asked for. But insisting that this is the only way that every requirement has to be framed creates unnecessary problems and risks and makes the team’s work harder than it has to be.

Don’t Get Hung Up on User Stories – Do Whatever Works

Insisting that all of the work that every development team needs to do has to be defined in the same way is arbitrary, unnecessary, and wrong. It's like back in the UML days when we learned that requirements had to be captured through Use Cases. Anyone remember how silly - and pointless - it was trying to model functional specifications using stick-men and bubbles?

Teams should use whatever format makes sense for their situation, with as little or as much detail as the problem and situation requires. There are times when I would much rather work with well-defined, detailed documented rules and scenarios that have been carefully thought through and reviewed, or a prototype that has been shown to customers and improved iteratively, than to be limited to a long list of 2-line User Stories and hope that I will be able to get someone who understands each requirement to fill the details in when the team needs them.

At Agile 2013, Jeff Patton in his talk on Agile Requirements and Product Management made it clear that the Story Template is a tool that should be used for beginners in order to learn how to ask questions. Like the “snow plow” technique for beginning skiers – something to be thrown aside once you know what you’re doing. His recommendation is to “use whatever you need to capture requirements: pictures, slides, notes, acceptance tests”. At the same conference, Scott Ambler reiterated that “stories are not enough. They are just one tool, a usage view. There are so many more options”.

Don’t worry about stories or epics and themes – or themes and epics (the Agile community can’t all agree on what’s a theme, what’s an epic, and it doesn't matter anyway). Add details when you need to. Get rid of details that you don’t need.

Don't get caught up in what Roman Pichler calls “Story Mania”:

Some product owners and teams are so fond of user stories that everything is expressed as a story. This either results in some rather odd stories – stories that capture the user interface design, complex user interactions, and technical requirements; or these aspects are simply overlooked.

Like any technique, user story writing has its strengths and limitations. I find stories particularly well suited to capture product functionality, and when applied properly, nonfunctional requirements. But user interface design and complex user interactions are better described by other techniques including design sketches, mock-ups, scenarios, and storyboards. Complement your user stories therefore with other techniques, and don’t feel obliged to only use stories.

Even some of the people who developed the original idea of User Stories agree that not everything can or should be done as a story.

Capture requirements in whatever format works best. Keep requirements as light and clear and natural as possible. If you have important technical requirements that need to be worked on, write them down and don’t worry about trying to warp them into some kind of arbitrary customer stories because you've been told that’s what you have to do in order to “be Agile”.

Let’s all stop telling stories and get some work done.


Greger said...

I've always thought that you don't need user stories but reading your rant was more fun than my more balanced view. :-)

Unknown said...

There are beautiful, simple ideas in today’s Agile development methods that work really well. And some that don’t. Like defining all of your requirements as User Stories.

You never were supposed to do this, particularly if (as seems to be the case) people thing this means “devein all your requirements only in user stories. Connextra certainly didn't do this.

Something significant and I'm tempted to say magical happens when requirements are put in the first person

That's the not the point at all. The point is to avoid (alleged) requirements of the form “the system shall…” which don't correspond to anything that anyone actually wants or needs to do—which was a common failure mode even of usecases. The idea of the role is that someone who does that role for a living told you a story about some value that accrues to them, and when you ned to know more you know what kind of person to go to.

Focusing exclusively on delivering Customer Value leaves little room for non-functional requirements and constraints,

This makes no sense. I agree that a lot of organisations fall down on this, but so–called “non–functional requirements“ are intimately connected with customer value. If the customer's business is time–critical then the performance of the system determines whether or not it can add value—a functionally correct solution that doesn't respond within the required time window (for example) odes not add value. This should be part of the conversation about the “I want…”.

And so on.

What you've done here is catalogue a bunch of ways that people mis–apply their poor understudying of stories. Which is a useful thing.

And it isn't their fault that people are getting into this mess—using stories they war they were intended is a technique that can be described in few words and then requires a lot of hands–on practice to apply effectively. That doesn't play well with the model we have now of “Agile” being a product that you buy.

The really important point here is the one made by Jeff Patton: these are transitional practices. Agile teams are meant to inspect and adapt, and if they are doing the same practices the same way after a year, or five years, as when they started, then they have missed the point. That doesn't fit well in the world of buying in a certified Agile process either. So it goes.

Anonymous said...

This is a very interesting discussion on the topic, and it underlies other issues with Agile implementation.

Commented on

Anonymous said...

So what you are saying is that you want to continue to develop "real" software using RUP or waterfall.

That's cool.

Don't conflate agile or scrum with RUP or waterfall. That's all I ask.

User stories aren't requirements. They are a way to map how a user might use your product. If you're still developing software with all of the requirements are handed off to you upfront, then cool.

But that isn't agile.

Jim Bird said...


Don't conflate Agile with stories. That's all I ask. You can be Agile (or at least small "a" agile) without writing stories. There are times when more detailed requirements are necessary, and most people would be better off if they didn't try to jam every story into a Connextra template when they do write stories.

Site Meter