A Project Management Playbook for Data Teams

A data team that's everyone's on-call analyst

Picture a data team of four to seven people, split between building and maintaining the actual data platform, pipelines, warehouse, dashboards, and answering the constant stream of requests that arrive because they're the only team who can query the data directly. Marketing wants a campaign performance number. Finance wants a custom report. A product manager wants "just a quick pull" that turns out not to be quick at all.

Each request is small and reasonable in isolation. Together, they're an unbounded queue with no natural limit, because unlike a feature backlog, there's no shared sense across the company of how much data-team capacity a "quick" request actually consumes. The team's real platform work, the migration that's been "in progress" for two quarters, keeps getting bumped by whatever arrived most recently.

This is a familiar shape for data teams: genuinely valuable service work slowly crowding out the infrastructure investment that would make future requests faster to answer.

Before: ad hoc work has no edges

The core issue is that ad hoc requests arrive through every channel at once, a Slack DM, a hallway conversation, a reply-all email, and none of them pass through anything resembling a queue. Whoever gets asked directly just does it, because saying no to a specific person feels harder than absorbing one more small task. Multiply that by five people fielding requests independently, and the team has no shared picture of how much ad hoc load it's actually carrying in a given week.

The second issue is that the platform migration, the thing that would actually reduce future request volume by making self-service possible, never gets protected time. It's always the thing that can wait one more day, so it waits for months, and the team stays trapped answering the same category of question by hand indefinitely instead of building the thing that would let people answer it themselves.

The third issue is that stakeholders have no way to check status themselves. Every "is my report ready yet" is another interruption, on top of the requests already in the queue.

After: the platform project finally gets protected time

The honest version of what changes is that ad hoc work gets a visible edge instead of being unbounded. A full column is a legitimate reason to say "this has to wait until tomorrow," which used to be uncomfortable to say without something concrete backing it up. The platform migration, protected by its own column and limit, actually gets sustained attention instead of being perpetually the thing that can wait one more day.

Stakeholders still need things from the data team, that doesn't change. What changes is how many of those needs get resolved without an interruption: a status check answered through a plain-language query instead of a Slack message, a metric definition found on a wiki page instead of asked in a thread. The team ends up spending more of its time on the work that reduces future request volume, not just the requests themselves.