businessanalytics

KPI Dashboard Specification

Designs a KPI dashboard with metric definitions, data sources, visualisation types, and alert thresholds for a department.

Prompt

Act as a business intelligence analyst for a [industry] company. Design a KPI dashboard for the [department] team whose primary goal is [goal]. The target audience for this dashboard is [team leads/executives/analysts]. Create a structured specification that includes: (1) 8-12 KPIs grouped by category (leading indicators, lagging indicators, health metrics), (2) for each KPI provide a detailed table with: name, formula/calculation, data source, update frequency, target value, red/amber/green thresholds, and the best chart type to visualise it (for example, 'a sparkline for daily trend, a gauge for current vs target'), (3) a suggested dashboard layout describing a 2x4 or 3x3 grid arrangement, (4) 3 drill-down views that users might need, (5) 2 automated alerts with trigger conditions and notification channels. You must ensure each KPI answers a specific business question, and state that question. Avoid vanity metrics such as 'total page views' without context. Only include metrics the team can directly influence. Tone should be analytical and precise.

Why this prompt works

The 'each KPI answers a specific business question' rule is the secret weapon, and the requirement to state that question explicitly stops the output from being a list of metrics that look important but nobody actually uses. Grouping by leading, lagging, and health is also a small but meaningful structural choice: it forces the model to think about which metrics are predictive versus which are scoreboards, which most dashboard tools blur together. The red/amber/green thresholds and chart-type recommendation per KPI mean the spec is detailed enough to hand to a BI engineer without a follow-up meeting.

When to reach for it

You're building a new dashboard from scratch and want a brief that the BI team can implement without reverse-engineering your intent.
You're auditing an existing dashboard that's grown to 30+ widgets and want a structured way to slim it down to what matters.
You're a new department head and the team's existing dashboards don't reflect the priorities you've just set.
You're writing a board update and want the underlying KPIs documented in one place so the next quarter's update is fast.

How to customise it

The 'goal' field for the team is the single biggest input. 'Drive growth' produces a generic dashboard; 'reduce churn from 5% to 3% by end of year' produces a dashboard whose KPIs map to that one goal. For executive dashboards, ask the model to favour leading indicators over lagging ones, since executives need predictive signals rather than rear-view mirrors. If your team uses a specific BI tool (Looker, Metabase, Mode), name it in the input so the chart-type recommendations reference visualisations actually available there.

What good output looks like

A spec doc with three groups (leading, lagging, health) of KPIs, each with a table row covering name, formula, source, frequency, target, thresholds, and chart type. Then a layout recommendation, three drill-downs, and two automated alerts. Total length 1,500 to 2,500 words. The thresholds are starting points to recalibrate against your historical data; the layout suggestion is genuinely useful as input to your BI team.

Watch out for

Dashboards designed in one sitting tend to over-include. The model will generate 12 KPIs because that's what was asked for, even if 6 would be enough. After the first pass, ask 'cut this list to the 5 KPIs the team would consult daily, and move the rest to a separate strategic-review dashboard'. The default thresholds (red/amber/green) are heuristic, not based on your historical data; treat them as a starting position to recalibrate after a few weeks of real numbers.

KPIdashboardbusiness intelligencemetricsanalyticsChatGPT / Claude