At the beginning of the year our team – who look after the core websites in the Ubuntu ecosystem – decided to adjust our project delivery process and go full-on Scrum. This is hardly shocking news these days. Actually, you can find way more people writing about how they decided to ditch Scrum and what a terrible idea it was to begin with. But for us, it seemed like a good thing to try and so far the results are promising.
Happy Web Team sprinting on the 16.04 release day
What is Scrum?
Scrum was born as a framework to manage the development of complex products where the traditional project methodologies failed. It has a very simple set of rules that are best explained by its founding fathers Ken Schwaber and Jeff Sutherland in the Scrum Guide. Unfortunately, over the years the simple framework has been – on the one hand – contaminated with various additional concepts and on the other, diluted by assumption that Scrum can work in any context or on a pick-and-choose basis.
When you think Scrum do you think user stories, burndown charts and planning poker?
Well, that is what I thought before I realised that all these things are NOT part of Scrum but the optional additions, which sometimes can help but sometimes can be detrimental.
Life before Scrum
I joined the Canonical Web Team at the end of last year in a project manager’s role. At that point, the team was experimenting with introducing some of the Scrum artifacts, like sprint planning and retrospective, into an already familiar Kanban method.
Standups were a daily bread for the team and kept everyone in the loop. Retrospectives made us reflect on the ways we work and how we can improve. Team members were skilled at planning their tasks and progressing through their work. The fact that we are a relatively small team based in the same space also helped with communication and collaboration.
What didn’t work
It felt like the partial adoption of Scrum was not really allowing us to reap all positive effects. We operated on a one week sprint basis. We had planning sessions at the start of the sprint but they were mostly focused on tasks for individual team members rather than team working out what we can complete each sprint and how to make it happen. We also didn’t involve the stakeholders in our meetings, so often team members didn’t have a good understanding of the business context behind the projects.
Our log of current and incoming work was massive, recorded in few different places, and often not very well defined. We work with multiple stakeholders from different functions and product lines. We also have a handful of internal projects aimed at improving our ways of working (for example a great Vanilla framework initiative you can read about here). It was a challenge for us to understand our short and long term priorities. It also felt that we can benefit if we make the stakeholders a part of the prioritisation and planning exercise.
Life after Scrum
Our team was one leg in the Scrum already and it intuitively felt that pushing it further could help us resolve at least some of the above mentioned problems.
We extended the duration of our sprints to two weeks in order to have enough time to complete usable product increments. Bonus – it reduced the overhead of weekly planning and review sessions.
We shifted our thinking about what we take on each sprint from the task level items to the backlog level items which ideally, should be demonstrable at the end of the sprint. The goal is also to get better at understanding our velocity, improve estimation and therefore be more accurate in planning what we can accomplish in two week’s timeframe.
This change created a more sustainable work pace while at the same time added a sense of urgency around most important work. Sprint clock is ticking, we know what we planned to complete and from day one we do our best to succeed. As mentioned by one of our developers, “Team now has bi-weekly rhythm and greater visibility of overall team goals in any given sprint. There is a constant march of progress”.
Introducing a backlog
Product backlog is one of the Scrum artifacts and our team didn’t really have one before. We decided to replace all docs and tools where we captured upcoming projects with one simple spreadsheet, hosting all briefs and other relevant project info. This allowed us to surface all the work that team had in the pipeline, including the internal improvements that didn’t have any owners outside of the Web Team. This new format supports us in defining and building the consensus about our priorities.
One additional shift related to our estimation process. We decided to discard task estimating altogether for the simplicity and focus only on the backlog items. The principle is that the higher the project is in the backlog, the more we break it down into small increments so our estimates are more accurate. We started with a big backlog refining session to get a better understanding of what we have in the pipeline. We now use the burnup chart which I update at the end of the sprint and it helps us to predict when, roughly, we may get to the projects that are further down the pipeline.
Meetings, meetings and few more meetings
We decided to introduce new meetings and slightly alter the existing ones. Every two weeks our Tuesday schedule is pretty hectic but it feels like it actually saves us time in the long run.
We start with a sprint review meeting for all team members and stakeholders. That’s where we look at what was planned for the sprint, demo what we completed and – if needed – explain what we didn’t manage to finish and why. We then move to the backlog planning session with the same group, agreeing what should make its way to the next sprint, clarifying outstanding questions and talking through new briefs that came in the meantime.
The second set of meetings involves only the Web Team. We first run our usual retrospective and after that, we plan the next sprint in details. We discuss all backlog items planned for the sprint, identify tasks needed to complete them and team members pick what they will be working on. After this exercise, even without estimating, we get a pretty good idea whether we’re overloaded or we need to add more work.
So far the review and planning meeting with the business turned out well and helped to improve the communication, understanding of everyone’s work, and consensus about the priorities. As one of team members pointed out, “stakeholders now have to agree to work beforehand and any ad-hoc work inserted into the iteration visibly impacts on our initial commitment”.
Scrum isn’t always a walk in the park
The beauty of Scrum is that it’s lightweight and in principle, simple to understand. It doesn’t mean however that running projects with Scrum is not difficult. Below I outlined what I see currently as our key challenges.
Launch deadlines vs. sprint schedule
Some of our web projects have strict launch dates as they are tied to specific events, such as a launch of new device. More often than not, these fall somewhere within the sprint duration so we can’t simply finish them at the end of the sprint. One way of addressing it would be to schedule them in the preceding sprint however we are rarely able to gather all necessary bits of information beforehand. For now, we just accept the fact that some items in our sprint have a higher priority and need to be released before the sprint ends.
Sprint planning and estimates
More often than not, by the end of the sprint we realise that we won’t complete all scheduled backlog items. We try not to make a big drama out of it – apparently it happens roughly half of the time even for mature Scrum teams – but at times, it deflates a sense of accomplishment. We can definitely improve how we break our backlog items, and our estimates are still sometimes based on a wishful thinking. There is also too much temptation to add more to a backlog than our velocity predicts – we try to resist it, with mixed effects.
Getting stakeholders on board
Introducing stakeholders to our planning and review process was seen by the whole team as a huge improvement, yet working with clients – either internal or external – is not always a piece of cake. We sometimes struggle to get good briefs and all necessary information before the sprint starts, and every now and then we get a cheeky request to fit something new in. It’s a long process to get everyone follow the rules and we are only at the beginning!
Bugs and business as usual
This seems to be quite a common problem for many Scrum teams – how should we deal with bugs and small requests (like copy changes) that come during the duration of the sprint? We considered postponing them until the next planning meeting where we would determine which ones we add into a sprint. Eventually, we decided we would spend more time discussing and planning these things than it actually takes to resolve them. Therefore, we assume that a small percentage of team’s velocity (about 10%) will be spent on dealing with these. That’s not quite in line with Scrum principles and we don’t have a good way to measure yet what’s the actual impact of such work.
Happy Scrum, everyone!
Scrum doesn’t work everywhere and for everyone but our team is happy with the direction we moved to, which often comes back in our retrospectives. Empowerment, openness, collaboration, and better structure have been mentioned by several team members. It’s really remarkable how easily everyone adjusted to the new process and team’s commitment to make the whole thing work has been invaluable.
We are only in our seventh sprint now and are still on the steep learning curve, making adjustments as needed. Like anything in this world, Scrum is not perfect but maybe the greatest thing about it is that it keeps continuous adaptation at its core so within each cycle, we can look back and change what we don’t like.
Want to learn how to be a Scrummaster or Product Owner yourself?
Before we started our transition to Scrum I completed the Scrum Master and Product Owner training classes with Martine Davos at Skills Matter in London, UK, which I highly recommend for any Scrum newbies.