A Project Management Playbook for Distributed Engineering Teams

Three regions, one codebase, a lot of edges

Consider an engineering organization of around forty people, split into squads across three regional offices, roughly eight hours apart from each other. The codebase was shared and the product was one product, but the working day wasn't. Each region's squad had a real overlap window with only one neighboring region, never with the third, which meant every piece of work that needed to cross a region boundary had to survive at least one handoff where the two people involved weren't online at the same time.

The handoff itself was the recurring point of friction. A squad in one region would finish their day partway through a piece of work, leave a comment or a message about where things stood, and the next region would pick it up hours later, working from whatever context had been written down, which varied a lot depending on how rushed the person had been at the end of their day.

Squad leads in each region ran their own version of sprint tracking, loosely aligned but not really unified, so a genuinely cross-region view of "how's the sprint actually going" required someone to manually stitch together three separate pictures.

Before: a handoff that depended on how good someone's end-of-day notes were

The clearest cost showed up as stalled work at region boundaries. A pull request opened at the end of one region's day would sit unreviewed until the next region came online, which was expected and fine, except the board often didn't reflect that the work was actually done and just waiting on review. It looked stuck, when really it was just waiting on the clock.

Cycle time, if anyone had tried to measure it consistently across all three regions, would have been dominated by these handoff gaps rather than by actual working time, but nobody was measuring it consistently, because each region kept its own numbers in its own way, if at all. That made a genuine cross-region comparison, or even a genuine cross-region total, hard to trust.

And a squad blocked on a decision from another region had no reliable way to flag it as urgent in a way that would actually reach someone before their overlap window closed, so blockers often waited a full cycle, sometimes sixteen or twenty hours, before getting real attention.

After: handoffs that looked the same regardless of who wrote them

Work still crossed region boundaries constantly, that's inherent to a codebase shared across three offices eight hours apart, and no tool changes that. What changed was that the state of a handoff stopped depending on how thorough someone's end-of-day notes happened to be. A pull request sitting open told the next region exactly what it was, done and waiting, without needing a message to explain it.

Cross-region cycle time became a real, trustworthy number instead of three separately kept estimates that didn't quite add up. And a genuinely blocked squad had a way to surface that clearly enough that the next region saw it first thing, rather than the blocker sitting quietly for most of a working day before anyone noticed.