gtm-dashboard-spec

GTM Dashboard Spec

A dashboard is a set of questions with numbers attached. Specify the questions first; tools and charts come last.

Step 1: Identify the audience and cadence

Ask who looks at this and when. Different audiences get different dashboards , never one dashboard for all three:

• Operator (daily/weekly): what needs action this week
• Leadership (weekly/monthly): is the machine working, where's the constraint
• Board (quarterly): trajectory, efficiency, predictability

Step 2: Write the questions before the metrics

5–7 questions the audience actually argues about. Examples by audience:

• Operator: "Which campaigns created qualified pipeline this month?" "Where do deals stall?" "What's outbound reply rate by segment this week?"
• Leadership: "Is pipeline coverage sufficient for next quarter's target?" "What's CAC payback by channel?" "Is win rate moving, and why?"
• Board: "Is net-new ARR growth efficient (burn multiple, magic number)?" "How predictable is the pipeline (forecast vs. actual, last 4 quarters)?"

Every metric on the dashboard must serve one of the questions. A metric with no question is decoration, cut it.

Step 3: Define every metric like an engineer

For each metric produce a definition block. Ambiguity here is why teams argue about numbers instead of decisions:

**Metric**: Qualified pipeline created
**Formula**: Σ amount of opportunities created in period where stage ≥ <qualification stage>
**Source**: <CRM> · object: Opportunity · fields: amount, created_date, stage
**Filters/exclusions**: renewals excluded, test accounts excluded, currency normalized
**Owner**: <who fixes it when it looks wrong>
**Target/benchmark**: <number, or "baseline TBD after 1 quarter">

Force decisions on the classic ambiguities: MQL definition, pipeline stage that counts as "qualified", attribution model (pick one, name its known bias, move on), whether expansion counts, date basis (created vs. closed).

Step 4: Specify layout

Top-left is the most valuable pixel. Order: the one number that summarizes health → the questions in priority order → supporting breakdowns. Max ~12 tiles; a dashboard that scrolls twice is a report, not a dashboard.

For each tile: question it answers, metric(s), chart type (default: big number + trend line; bar for comparisons; table for "which ones" questions , avoid pies), time grain, comparison (vs. target, vs. prior period).

Step 5: Name the data gaps

List every metric the current stack can't produce cleanly, the missing piece (field, integration, hygiene rule), and the workaround until it's fixed. A spec that pretends the data is clean gets abandoned in week 2.

Output

Deliver as one document: audience + cadence, questions, metric definition blocks, tile-by-tile layout, data gaps, and a build checklist ordered so the top-left tile ships first.

Related skills

campaign-brief-builder for the measurement plan of one bet. landing-page-teardown and outbound-sequence-writer for the events worth instrumenting.