There is a certain connotation with the word estimate. People think of cost and time. Think about the last time a mechanic fixed your car or you hired a painter to put a fresh coat of paint on the third floor windows. You are thinking about time and cost aren’t you? When we start thinking about a software project we are still estimating the cost and time, but of the project not the stories.We are doing ourselves a disservice when we say we estimate our stories.
We have been relatively sizing stories on projects for years. We are not estimating our stories. When we look at a group of stories it is pretty easy to compare them relatively to each other and have a good understanding of which ones are similar. The hard part is to predict how fast we will finish each story. It is the velocity at which we complete stories, combined with the total number of points, which allows us to estimate the project’s cost and length.
So why does it matter if we size our stories or estimate them?
As soon as we talk about estimating stories the client (and team) naturally start to think about the time it will take to complete a story and how much that story costs. This leads to people wanting to change the number of points associated to a story because “it is taking longer than the estimate”. Just because a story has taken longer than anticipated, does that mean its relative size is different to all the other stories? Maybe… but most often it is that our velocity is not what we initially thought it would be. Talking about the size of the story helps to focus on the velocity being the value that is different from expectations rather than the number of points on the story. With this mind set, we can move away from constantly updating the the number of points on every story and instead focus on improving our velocity by finding ways to be more effective and reduce waste.
Learning a new concept or process works best when done in a simple situation. Once you have mastered the process, you can apply it to more complex situations. I’ve made the mistake before of introducing relative sizing and velocity calibration to a group with the actual stories that were going to be used for the project. Though we’ve managed, there has always been a challenge of re-explaining the process throughout and pausing to discuss edge cases.
In a recent inception, I decided to introduce the concept first with an abstract example… fruit. I had worked this through in my head, and was happy that it would introduce the basic concepts and make the actual relative sizing and velocity calibration much easier. I was right on that sense, but it also did much more. We were able to see many real world examples and pitfalls come to light which made the actual relative sizing and velocity calibration go much easier.
To start with I had written about 30 different pieces of fruit down on story cards. Rather than relatively size the stories based on the effort it would take to develop them, we were going to relatively size the fruit based on the effort to eat them!
Here are some of the lessons learned while doing the relative sizing:
The Technical Story
Things were moving along well and the team was making good progress through the sizing of the fruit. Then came the kiwi… someone commented, “we need a knife”. I asked them what they should do. The team wanted to raise a story to buy a knife. I asked what would happen if they played the knife story then the business decided not to play the kiwi story. They quickly agreed that if we never played the kiwi, buying the knife would be a waste. This led us to making an assumption that the kiwi story would include buying a knife, hence more effort.
When we later hit the pineapple story, we assumed that we already had the knife. There was a brief discussion about what happens if the kiwi is deprioritised and the pineapple is not, and everyone was comfortable that the assumption of already having the knife would then be wrong and we would resize.
- Do not create stories that will introduce technical infrastructure
- Include the effort to do this in the first story that needs it
- If multiple stories need it, make sure to assume the one that will be played first and assume in the others that it is already in place.
- If those assumptions change, you resize the stories.
The Unknown Story
There were two other fruits that caused us trouble. The jackfruit and the dragonfruit were both unknown to the team. The jackfruit was easier to resolve. One of the team members knew what it was and was able to describe it in enough detail for the team to size it relatively to the other fruit. The dragonfruit was much harder (I was the only one who knew what it was). Soon someone had an iPhone out and had it up on wikipedia.
- The team should support each other by sharing knowledge about the stories, allowing them to come to an agreement on the size of the story
- When the team doesn’t know the answer immediately, the BA should go away to discover the information.
- Set the unknown story aside until the information is found
- Capture any assumptions that you make
The Duplicate Story
Though we didn’t do this at the time, I’ve since realised that you could very easily include duplicate stories in the exercise. Depending on where you are from and where you have travelled, finding fruits with multiple names may be a challenge. Two examples are: carambola & starfruit and chikoo & sapodilla. This would be a nice way to set everyone’s expectations that there will be duplicates and to keep an eye out for them.
After the relative sizing, we moved on to velocity calibration. We decided on a team size of two pairs and decided that we would have 4 iterations of 15 minutes each. We asked the team to select the different fruits that they felt they could eat each iteration (with the assumption that they had a chance to digest their fruit between iterations). Besides learning the basic concept, we also encountered what happens when you try to include your Too Big stories.
The Too Big Story
As we were doing the velocity calibration, we had one iteration whose velocity was quite a bit bigger than the others. No prizes for guessing which fruit was included in that iteration. After a bit of a discussion it was easy to see that we shouldn’t include the watermelon in the velocity exercise. In fact we realised that the watermelon could be broken down into smaller pieces – think of the number of times where you have seen a half or quarter of a watermelon for sale in the produce section…
- Exclude your Too Big stories from velocity calibration
- Break your Too Big stories down, resize them and include the smaller stories
When we repeated the exercise with the real stories. The BAs and SMEs were aware that they needed to answer questions on the unknown stories, we excluded the Too Big stories from the velocity calibration, and the team was diligent at capturing assumptions. Numerous times I heard someone comment “we need a knife for that story”!
Most importantly though, the entire team had done the fruit exercise. SMEs, PMs, BAs, QAs and Devs were familiar with the process before going into the actual sizing and calibration. For the devs the benefit was obvious, but the understanding from the other members of the team was invaluable. Even though the other team members recognized that they could no longer participate with the exercises, they were engaged and involved throughout. They took turns facilitating, answering questions when they could and rarely challenged what the developers decided!
This is often a question that comes up on a new project. Folks who have been doing iterative development for a while rarely want more than 2 weeks and often only 1 week per iteration. Yet folks new to the game push for more time; 4 weeks per iteration is quite common. The most common reason that I hear for wanting longer iterations is that 1 or 2 week iterations means that too much time is spent in showcases and retrospectives. Which leads to an interesting question, what is the goal of an iteration?
This is one of those questions that if you ask 10 people you will probably get 11 answers. For me, an iteration is nothing more than a heartbeat for the progress of the project. I’m not worried which stories we complete in which iteration, I’m not worried if we have a showcase or a retrospective. What I want from iterations is a frequent measure of our progress towards our end goal, releasing software that delights our customer. Of course release planning, retrospectives and showcases are an important part of delivering software that makes our customer happy; they may just be needed at a different frequency to your heartbeat of progress.
Once we separate the frequency of capturing your heartbeat from that of your planning meeting, showcase and retrospective, I find that the willingness to have shorter iterations is much greater. Great… but now we have four questions to answer:
- How often should we plan what should be worked on next?
- How often should we have a showcase?
- How often should we have a retrospective?
- How long should our iteration be?
My first project with ThoughtWorks 12 years ago had 2 day planning meetings. Talk about waste! That was two days of development that we lost out on, but we were new to iterative development and were learning! Now I never see a project that has a two day planning meeting, but I do see projects that have half a day or two hour planning meetings. That is still a lot of time away from development.
Rather than having a planning meeting every iteration, why not pick the next most important thing to work on when you are just about* ready to need something new to work on? This allows us to be flexible when priorities change and avoid having the whole team locked up in a room for 2 hours. Plan continulously when needed rather than on a fixed schedule.
*Of course we want to make sure people are not sitting idle, so you need to find out what just about means for your team.
Let’s take the showcase next. Ideally as often as you can! Why wait a week or two weeks, if you have something that is ready to be seen by your stakeholder, pull her over to your desk and show it. Better yet, show it while it is still in progress and then again when it is finished. Frequent feedback cycles reduce our chance of waste. For showcases to a wider set of stakeholders with busy diaries, you need to find a regular time that you will have sufficient work done to make the showcase worth attending but not so infrequent that you are having to make numerous changes based on their feedback.
Retrospectives are easier, they are for the team. You don’t need to worry about busy stakeholders who are trying to support your project and do their day job at the same time. On well functioning teams you have constant retrospectives. The discussions on the way to lunch or at the pub after work are a great time to look at how we can improve. Unfortunately, it is difficult for everyone to be involved in those conversations and a formal retrospective is a great tool. Look at how your team is functioning and the value you are getting from the retro. Have them as often as you can as long as the value is still outweighing the cost.
Which brings us back to our original question. How long should your iteration be? The explanation that I’ve used recently is “frequent enough to avoid missing crucial information but to not cause huge fluctuation in points from iteration to iteration”. Too frequent and we lose the ability to predict what we will achieve in the next iteration. Too infrequent and we can realise too late that we are off course.
Stop thinking that an Iteration has to be made up of a pre-definied set of stories, have a planning meeting, showcase and retrospective. Measure at the right frequency, plan continuously, showcase when you have something to show and retrospect as needed. All four activities are valuable, maximise the value they bring to delighting your customer by doing them at the frequency that is right for each activity.
Last night Mike and I traveled from Chester to London to see the Michael Vincent lecture at International Magic. It may seem like a long trip to see a lecture, especially knowing that we would be on the 7:10 train this morning to get back to work, but Michael is a wonderful lecturer and great magician. Unfortunately there were signalling problems on the lines which caused delays to our journey and resulted in us missing the first half of the lecture. The way the delays were handled led to an interesting discussion about sharing information and the correlations to delivering software.
We heard a rumour when we got on the train that there were signaling problems, but being that there was no announcement, we assumed that things were fine. We pulled out of Chester on schedule and only just before arriving in Crewe was the first announcement made. It looked as though the delays were going to be quite severe. We were about to head back to Chester when a second announcement was made: there was a train leaving now for London that would take an hour and forty minutes. We did the math and saw that we would only miss a few minutes of the lecture. We ran to the platform, found seats and soon were under way. As the train pulled out of the station, the train manager announced that due to the signaling problems they could not estimate when we would arrive in London. It was too late, we were committed to the journey!
Mike and I started discussing how this was handled and the similarities that can be drawn to delivering software. The staff were all very pleasant and apologetic, but there had been one glaring mistake that was made… they waited too long to share information with the passengers. Why was this?
Quite possibly because they made an assumption that everyone on the train had the same goal: to travel to London before the next day. If that assumption was true, then a delay of an hour, two hours or even 4 hours wouldn’t matter. We would still all make it to London and it was better than waiting until the next day. The problem is that wasn’t the goal for everyone on the train, Mike and I were traveling for the lecture. If we arrived in London too late, the journey would be of no benefit and would have cost us time and money. We often see a similar situation in software delivery, the importance of time to market. If a piece of software takes too long, it may not provide the benefits that were originally set out to achieve. The right decision may be to invest the time and money in something else.
Who is at fault here? The delivery team? The business? Both? Neither? It doesn’t really matter, the important thing is to avoid situations where decisions are made too late and by the wrong people. It is a two way street and to have the best chance of a successful engagement, all parties need to share information. Stakeholders, by sharing your true goals behind the project, the delivery team is better equipped to know what information is important to you and when you need it. Delivery team, if in doubt share information and share it early. Allow your stakeholders to make the decisions that they need to make and understand when it might not be what you hoped. In the long run the relationship will be better off for it, and there will be many successful deliveries (they may just not be what you originally planned).
One of my biggest frustrations on a project is when we have stories that capture a solution rather than a goal. Often this leads to long conversations with the business to drive out exactly what it is that should be built and hinders the creativity of the development team in finding the best solution for the problem. I’ve spent a lot of time over the last year and half thinking about this and trying out a solution to the problem and now after much encouragement from Mike & David I am finally blogging about it. This isn’t meant to be a silver bullet solution, but I hope it provokes some discussion and you find it interesting.
A few of us got talking while doing an inception early in 2009 that the ‘so that’ statement was often either repeating the requirement, was too general or was left off completely when capturing stories. This meant that we were missing the true goal of the business and lead to problems with scope and misunderstandings of what a story truly meant once we got into delivery.
Using an example of changing channels on a tv with a remote control, let’s look at some of the poorly written headlines you could have…
Headlines which repeat the requirement as the business value tend to be very difficult to prioritise because we don’t understand the benefit that they provide to the user.
As a viewer I want to change channels on the tv with my remote so that the I can use the remote to change channels.
Headlines which are vague cause debate over alternative solutions that may or may not solve the problem. Is using the remote really necessary to watch different channels?
As a viewer I want to change channels on the tv with my remote so that I can watch different channels.
Finally, in really bad cases you end up with no business value at all which are both difficult to identify alternative solutions for and to prioritise:
As a viewer I want to change channels on the tv with my remote control.
Of course there are always the times when a headline captures the true value of what the business is trying to achieve:
As a viewer I want to change channels on the tv with my remote control so that I can watch different channels without needing to be in reach of the television.
This story can be easily prioritised against other stories and alternative solutions identified, but there are still problems. The headline ends up being quite long winded and you may need to read it two or three times to grasp what we are aiming to achieve.
We played around with the idea of switching the so that and the I want but weren’t happy because it was very awkward when reading it. We got busy with the inception and the pressures of delivery overtook.
My next project was centered around helping the client learn to write good stories and develop their agile process. I had lots of time to think about what goes into a story and what is most important. My favourite question came back… how do we ensure that the true goal of the business is the most important thing captured in a headline. I came up with a potential solution, but didn’t teach it at the time because I hadn’t tried it myself. I’ve now used it successfully on my current client for almost a year and found that it has accomplished what I wanted and more.
The format that I use is:
As a (role) I want (business value) by (method/requirement)
This forces the writer to capture the value and be specific with what what the story is aiming to achieve. At first I found it quite challenging, but soon realised that the difficulty wasn’t because of the format, but that it was forcing me to have headlines which clearly stated the goal the business was trying to achieve. I wasn’t allowed to wimp out! The feedback that I have received from the developers on the team has been very positive and it has helped to ensure a common understanding with the business.Our remote control example now reads:
As a viewer I want to change channels without being in reach of the tv by using my remote control.
This states the same benefit to the viewer and solution as the ‘good’ headline above, but in a much clearer and natural statement. We can prioritise it easily and the team has the ability to think of alternative solutions. It also makes it very difficult to run into the ‘bad’ headlines from above.
Once I started using the format, I realised there was another big advantage. The business value is attached to the strongest words of the statement (‘I want’) and the method of delivering the business value is attached to the weaker word (‘By’). In the traditional I want/so that format it is the other way around. This makes a big difference in the conversations with the business when an alternative solution has been identified, but that is for a different post…