Home / Work / More stories

Atlassian

Three shorter stories about influence, simplicity and stopping

Not every important call is a launch. These are about getting something built with no team, shipping less, and pausing work customers wanted.

App switcher

I shipped a product with no team, by making it serve other teams' goals

Early at Atlassian I was given two things: a shared role on a high-priority navigation overhaul, alongside a Principal PM, and sole ownership of the app switcher. The switcher was an orphan. It had no team, and every product had added its own features to it over the years, with nobody setting direction.

I saw it differently. It was the one place across every Atlassian product where people land when they want to move between them. Get it right, and it wasn't a tidy-up. It was the most powerful cross-sell surface in the company. But every engineer was busy with the navigation overhaul.

Do the work first

With no designer, I made the mockups myself in PowerPoint. The platform's lead designer saw what I was doing and volunteered to help.

Reframe their metrics

The growth team was already experimenting in the switcher, measuring each product on its own. I showed them a unified switcher would let them test across products at once, with cleaner results. They gave it two engineers.

One component for everyone

We shipped a single switcher any product could adopt, starting with high-traffic Trello.

No. 1Cross-sell touchpointAcross the whole Atlassian suite
1 of 4Platform essentials in FY20Every product was required to adopt it
The Atlassian app switcher open over Jira, listing every product on the site and a recommendation for the team
The app switcher today: every product on the site in one place, with room to recommend what the team doesn't have yet.

They didn't need to believe in my vision. They needed to see it served their own goals. With no authority, that is the only currency you have.

Sprint reports

Shipping one report instead of two delivered more

In 2019 Jira was bringing reports to its newer team-managed projects, built for teams who found classic Jira hard to use and were trying simpler tools. Two well-known reports were missing: the burndown chart and the sprint report. The safe option, which my manager favoured at first, was to rebuild both, because customers were used to having two.

The Reports overview for a next-gen project, with the new Sprint burndown chart and Cumulative flow diagram
The reports page for next-gen projects, October 2019. The Sprint burndown chart and the Cumulative flow diagram were new.

Why one report, not two

We spoke with more than ten customers and watched how they used both. It was often the same people, at different moments. During the sprint, they used the burndown chart in stand-ups to spot scope changes and remove blockers. At the end, they used the sprint report's table of what was and wasn't done for retrospectives and for reporting to stakeholders.

Both reports answered the same question: what might stop us meeting our sprint commitment, and how do we get better next time? Keeping two meant more to learn for every new user and more for us to maintain. So we took the best of each, the chart from one and the table from the other, and made them a single report.

The same people, two moments

Stand-ups during the sprint, and retrospectives and stakeholder updates at the end.

Duplicated work

Two reports to learn and two to maintain, both answering one question.

Quality, not removal

Framed as raising the bar, not taking a report away, which is how we won the argument.

What we reimagined

The Sprint burndown chart: remaining work against a guideline, with a tooltip for each scope change, and a scope changes log below
The chart shows every scope change in place. Below it, a log of what changed and when.

A scope changes log

Every issue added or removed, and every estimate changed, after the sprint began, so a team can see exactly what knocked it off course.

Sort by epic and assignee

New columns for epics and assignees, sortable, so a lead can see progress on the most important work and unblock people on it.

A fresh design

Rebuilt with Atlassian's design system, with an honest summary of how much the commitment changed.

Sorting the table by epic, status and assignee.
A scope changes log with a summary: the sprint commitment increased by 3 story points, with 46 left to complete
The summary at the bottom says what the changes add up to: here, 3 more story points.

What happened

After it shipped, use of the sprint report and the frequency of burndown use both rose, and customers asked for more reports rather than the old one back. Feedback on the launch post called it more helpful than the regular burndown. Other comments asked for the next things: per-person burndowns, non-working days and a control chart.

The engineering manager thanked us for pushing for something the team could be proud of, and our Head of Product later used it as an example of shipping from needs rather than legacy structure.

Multiple boards

Pausing a year of work was the right call. Resetting expectations was the hard part.

I led the work to bring multiple and cross-project boards to team-managed projects, something most enterprise customers relied on. It was public and in demand: the request had more than 1,700 votes, and we had invested close to a year.

As the work went deeper, it became clear how much was still needed to reach an enterprise quality bar. At the same time, large customers were feeling performance and scale problems that competed for the same engineers. Both directions were valid. Rather than let the project drift, I set out the remaining effort and the opportunity cost for leadership as an explicit choice: keep investing in a narrow but visible feature, or move the team to foundations that would help far more customers.

Paused, not killed

We committed to reassess the decision, not declare it dead.

Told customers myself

With support and design, I explained the change to early-access customers and on the public request.

Kept what we learned

It fed the boards guide, my Team '22 talk, and work on other products with multiple boards.

The team moved on to unifying the board and backlog and to performance at scale. What I'd do differently is write down, at the start, which assumptions would trigger a stop-or-continue decision, so the conversation is about criteria, not whether the work was worth it.