Developer Profile
An individual developer profile reached by drilling in from a team view, answering one question for an engineering manager: how engaged is this developer, and where has their effort been invested?
- CEO intake · Jun 1
- 5-persona audit
- 22 logged decisions
Origin
Intake was a meeting with the CEO, with the business stakes stated bluntly: a prior enterprise prospect had been lost, and the CEO tied this page directly to that failure: "We only had metrics dashboards and goals. Very, very underwhelming." The design brief in one line: engagement, not hours. The audience is the engineering manager (explicitly not the developer self-reviewing), and the page behaves as a drill-down, never a standalone destination.
Not so much busy in terms of hours, but how engaged is the team.
Data grounding, and a self-correction
The data-relevance pass surfaced the stream's binding constraint: activity is captured finely at the edge, but stored facts collapse to a daily grain in a fixed timezone, so the times-of-day grid every stakeholder wanted cannot be built without new pipeline work. It also caught its own earlier mistake: a first pass graded the human/AI code split as measured; a later validated pass showed it is estimated. The downgrade was logged in the objectives and data docs rather than quietly absorbed. The paper trail is the point.
Design decisions
Blocked, not faked.The times-of-day grid stayed in the design, drawn empty, stamped "DATA BLOCKED," with all unserved asks later consolidated into one visible "coming when the data lands" strip instead of scattered empty panels.
Engagement over output. The activity heatmap defaults to active minutes (an engagement signal) rather than commit volume, an output signal that is trivially inflatable.
Honest grouping."Project = repository" for v1, the only fully-populated, non-overlapping grouping available; work-area allocation stays a labelled secondary view with an overlap caveat.
Proxies get benched. Focus-block and context-switch metrics are proxies, so they were deliberately kept out of the headline.
The persona audit
The only stream to face a full five-persona panel: CTO, Dev Manager, Dev Lead, Developer, and UX Designer. Eight consensus findings; seven implemented as numbered decisions, five escalated as open calls rather than invented. The dominant problem: the clean stakeholder view had stripped audience-critical qualifiers into the hidden review layer; four of five reviewers independently flagged two work-area bars summing to 96% as reading like a bug. The second cluster: a hidden, gameable "power-user" threshold that incentivized token-burning, the opposite of the cost narrative it served. What all five praised: the blocked-data panel, and the refusal to imply "spend avoided."
5
Personas on panel
8
Consensus findings
7
Implemented
5
Escalated as open calls