Dashboard Designer

A dashboard without a job becomes a data museum. The first question the skill asks is whose desk this lives on and what decision it triggers — founder relaxing or worrying, ops team unblocking work, board confirming progress. Different jobs need different dashboards. A single dashboard trying to serve all three serves none of them, so the skill will recommend splitting before it recommends adding another widget.

The VITAL framework picks the metrics. Volume (how much activity), Intensity (how deep the engagement), Throughput (how efficiently it converts), Alarm (what would signal a problem), Leading (what predicts the next quarter). One to two metrics per category, nine total maximum. If you can't get below twelve, that's the signal to split into two dashboards rather than cram more onto one. Anti-metrics get listed explicitly — the things you chose not to track, with the reason — so the dashboard doesn't bloat back to forty charts the next time someone says "could we also add..."

Industry starter packs exist for SaaS, e-commerce, agency/services, marketplace, and content/media. Each gives a five-metric default that maps to VITAL. The starter pack is a starting point, not the answer — the skill customises against the actual business stage, because a pre-PMF startup tracking MRR growth as the headline metric is doing performance theatre rather than performance management. Pre-PMF gets engagement and retention only. Scale-stage gets cohort analysis. Stage matches dashboard complexity.

Alert thresholds use a hierarchy. Historical data first — green is within one standard deviation of the trailing six-month average, amber is one to two, red is beyond. No historical data? Use the industry benchmark from the starter pack. No benchmark? Use the business's own stated target. Nothing exists? Set thresholds based on the level at which the metric starts costing money or losing customers. Every alert has a specified action and timeframe — "review within 48 hours," not just "look at it."

You get a dashboard spec: the purpose line, audience, cadence, the scorecard metrics in a table with sources and alert thresholds, the layout broken into scorecard row, trend row, and segment row, a chart-by-chart description of what each visualisation shows and what it deliberately doesn't, the alert rules with named owners and response times, and the anti-metrics list. If you can't identify data sources for proposed metrics, the skill flags it as an implementation blocker rather than designing around data that doesn't yet exist. A dashboard without data is a wireframe.