We had a lot of problems trying to use user stories as the basis for a product backlog.  We haven't had a retrospective yet, and I anticipate having to wear my asbestos boxers for that wondrous event; however, I have noticed that some of the bigger picture aspects of what I tried to accomplish have established roots in the team.

Phrasing Requirements

One of the things I kept beating into the team's head was Mike Cohn's user story template: 

"As a <user role>, I want to <feature activity> so I can <value achieved>."

Sticking to this template helped us in many ways:

• it pushed the team into thinking from the user's perspective, which greatly improved our ability to communicate about the feature by having a common, iconic user to reference;
• it forced a value statement for every feature, which made blatantly evident the features that were also-rans;
• it made creating acceptance tests a breeze.

Several other projects have flared up at the office, and what I'm seeing is the product owners continuing to use this template to communicate their requirements.  That means better discussion and more acceptance testing.  That makes me happy.

The Need for Accountability

One of the points of using user stories as the product backlog was to enable the team to develop vertical slices through the system - get a single feature working end-to-end - rather than horizontal slices - building up an entire system layer, such as a database or front-end in isolation.  I wanted a cycle of feedback for the team, get a feature into QA for testing, in front of the stakeholders for comment, back to the developers for refinement.  The softer side of the team - by that I mean the product owner and customer representatives - wanted that too.  But the vertical development never happened.  The developers simply would have no part of it, citing various reasons that are locally valid but miss the global point.  Instead, they chose to work on horizontal layers of the system.  Milestones came and went, no features were evident, and QA was sitting on their hands for over three weeks.

The stakeholders brought up the notion of recourse: how would development be held accountable for the missed milestones?  The answer, much to my chagrin, was that they would not.  There is no delicate way to explain the situation, but it may help to understand that the efforts of the developers were praised by their senior manager.  

I've been reeling over this for a week now, my eyes bloodshot from searching for a positive in this dim and ugly situation, and the thing I keep coming back to is this notion of transparency.  Everyone - the team, the stakeholders, and the innocent by-standers - knows why milestones were missed.  So at least we can collectively acknowledge the situation, and that there is a need for holding the team accountable for disrupting the feedback loop.

Or not, depending on the way the wind blows... 

As recounted here, I had the team circle the wagons and pump out a bunch of user stories for an integration project.  It was the first time we'd tried user stories as a means to express requirements, but things felt solid and complete, and the product backlog was full of stories organized by system component and user role, perfect for developing in vertical slices. 

The project was considered top-priority and had the undivided attention of the company for the time being.  In the rush, I skipped estimating the backlog items and scheduled the sprint kickoff meetings.

Over the product backlog meeting, we iterated through each user story, reading it aloud and discussing in enough detail so everyone understood the feature and the value it brings to the user, but leaving out implementation discussions.  At least, that was my expectation.  The team members uninitiated in the project got flustered with the lack of specificity in the stories.  Some team members seemed to panic, like we had missed a big part of the picture; others got frustrated at the level of intimidation in the room.

The sprint kickoff was a bazillion times worse.  During the product backlog meeting, the product owner hadn't expressed a desire to see a specific set of user stories first, so individuals took the sprint kickoff in their own direction and were adamant to work on features that were obsoleted or purposefully left out of the product backlog.  There was no common ground to be found.  At the end of that sprint kickoff meeting the team had no goal, no milestones, no tasks ... we couldn't even set up the time for our daily scrum.

I've spent a lot of time decrypting what exactly happened.   There were things under my influence, and others outside of my control.  Here's where I think I failed, and what I've taken away from it....

Failure: The product backlog meeting was the first time the entire team had come together to discuss the project.  In particular, it was the first time any of the developers were exposed to the feature list.

Lesson: Expose representatives from each team to the relevant product backlog / user stories before the kickoff.

Lesson: Have the team estimate the complexity of each story, to elicit discussion if nothing else.  Ensure that representatives from development and testing are present.

This seems really obvious, but after a week of daily 3- to 4-hour story workshops I was feeling like the team was familiar with the stories.  Usually there is a lot more effort put on the product backlog than what we did for this project.  The product backlog should be treated like a pet requiring daily attention and affection to remain healthy, but instead we just chain it outside with an open bag of food.  Consider that the team would normally help the stakeholders evaluate priorities by providing rough estimates of the stories in the product backlog.  If one story has a high business value but would take 6 weeks of work, while 3 stories with a greater total value would take one week of work, it may make more sense to tackle the 3 stories in one week first.  Congealing the team on those estimates requires discussion of the features.  We didn't do this estimation because of time pressures - but in so doing (or, I guess, in so not doing), we neglected ourselves of that vital team conversation.

Failure: The user stories varied in specificity from hazy to anal; some overlapped each other; some were really just acceptance test criteria for another story.

Lesson: Focus each story on a user role, a feature, and a business value.  Don't repeat the combination in another story.

Lesson: Express detail and acceptance test criteria as story notes, not as separate stories.

Lesson: Spend the time to refactor the stories and organize the product backlog.

A lot of the frustration in the product backlog meeting stemmed from the fact that some features had a single story, while others had multiple stories, all of them nearly identical save one bit of detail or acceptance criteria.  The perception was that little to no thought had been given to those sparse stories.  I think most, but not all, of this frustration could have been avoided if I had worked with the product owner and customer representative to consolidate and organize the stories before bringing them to the team.

Failure: The sprint kickoff meeting started with a goal vacuum.

Lesson: Force the product owner to prioritize milestones for the sprint before the product backog meeting.

After all, that's the point of the product backlog meeting, right?  To get the team to commit to a specific set of goals delivered at a fixed time.  Not having that goal means the story buffet is open, and each team member will want to do what they think is the most important thing on that backlog.

And last in my list, but certainly not the last mistake I made...

Failure: While the testing, customer representative, and product owner understood the nature of this new user story beastie, the development team did not.

Lesson: If you expect the team to participate in a new thing, make sure they understand its nature.

It's completely reasonable for someone to get frustrated with a process if their expectations of the process haven't been managed.  If I had simply reiterated to the development team that these stories were really placeholders for conversations about the feature that we'd have during the sprint kickoff, it probably would have stemmed a lot of the frustration at the product backlog meeting.  

All in all, it isn't just vinegar; a lot of sugar came out of us trying user stores, which I'll explain in pt3: Two Steps Forward.

A few months ago I agreed to fill the role of acting Scrum master so I could try out some new techniques with the team.  One of the things I've been itching to try is utilizing user stories as the basis for a Product backlog.  I got the idea from a few of Mike Cohn's articles on his site, as well as his book which I highly recommend.  It seemed like a fantastic idea, for several reasons:

1. User stories are feature-centric and not implementation-centric.  That is, they describe a feature from a particular user role.  I firmly believe the Product backlog should be kept feature-centric as well, for reasons I may expound on some other time.
2. User stories are expressed in "business" language.  This makes them very easy for sponsors to understand, discuss, and even create.  It also helps keep the conversation focused on the value the story brings to the project.
3. User stores are simple and relatively terse.  Each one fits on a single index card.  That makes them very amenable to centralized backlog management tools, like Tackle.
4. User stories easily adapt as the understanding of the business need changes.  This is exactly what the Product backlog should do.
5. User stories are estimate-able and independent, just as a Product backlog entry should be.

Now, I should point out that I had never actually written or used user stories before, I've just read about them and heard anecdotes from my peers.  They've always seemed like a better approach to what I've experienced, which is the massive requirements collection that reads like home theater assembly instructions and is dated the moment it's written down so even with 80 pages of spec I need to walk over to the product owner to get questions answered.

The target project was an integration project with some new features on an established product.  There had been some ... adjustments ... in staffing and people were having to fill new roles.  There was a lot of confusion about how to drive the project and where to start.  In short, a perfect time to try something new...

Cutting Teeth in a StoryStorm

I had the customer representative, the product owner, and some members of the team commit to a four-hour storystorming session.  The basic idea is to get a bunch of people in a room for a timeboxed session where everyone cranks out as many stories as they can; the goal is breadth, not depth.

One person on the team had some exposure to user stories at another company, so I asked her to spend 15 minutes or so describing the concept to everyone.  I emphasized Mike Cohn's wisdom about phrasing each story from a specific user role, expressing a specific business feature and it's value to the user.  We discussed how the pile of note cards would become the product backlog for the project.  We walked through developing some example stories that were relevant to the project.  Everyone understood the process and thought it would produce a usable backlog.  So I passed out index cards and pens, set my egg timer for 5 minutes, and asked everyone to write as many stories as they could think of.

And they froze.  The ticking of the egg timer was deafening amidst the occasional tap of a pen on the table, the crackle of an index card being sacrificed to the Ultimate Outbox.  No one was writing a single thing.

After a few minutes I realized my mistake - it was completely unreasonable for me to expect the team to be comfortable cut loose on an open-form task they had never tried before.  They were afraid, and rightfully so.  They were filling new roles on a new project that was very different from the norm in a company that recently had lay offs.  I quickly thought back through the basics:

Software requirements are a communication problem.  User stories are one way of addressing that problem; each story is a placeholder for a conversation between the developers, testers, and stakeholders. 

So I killed the timer and suggested that we start talking.  We had already made a list of every user role in the project during the overview; to provide direction to the conversation I picked a single user role and we started discussions with the first thing that user would have to do to use the product.  It took some time to feel out how to hone the discussion into user stories, but following Mike's phrasing template was invaluable to accomplishing this.  After a while the index cards were getting pumped out faster than I could keep track of them, and the process started to feel natural and comfortable.

We finished that first session and did a mini-retrospective on the process.  The general consensus was positive, although some team members opted to withhold their joy until they see the project features implemented.  We scheduled several more sessions over the next week to round out the backlog.  Some sessions were spent standing in front of the product, actively modeling the user activities while a scribe jotted down the stories.  Overall, it seemed to be a very effective and productive way to produce the project requirements.

Little did I know that I had made several mistakes that would nearly derail the project...

... which I'll discuss in pt 2: the Kickoff that Didn't.