Chapter 2 — Practice-on-your-own solutions

The trap. The request looks like one artifact for one audience («a dashboard») but the head of support and a team lead make different decisions and need different comparisons. Merging them produces Marcus’s compromise dashboard: six visuals that fail both audiences. Splitting them produces two small charts that each answer a specific question.

BRD #1 — Head of Support (portfolio decision).

• Audience: Head of Support, weekly ops-review meeting.
• Decision: Which team(s) to re-staff or re-scope next planning cycle.
• Key question: Which teams are running above or below their sustainable ticket-load capacity, and by how much?
• Comparison: Actual ticket-load per team vs. that team’s sustainable-capacity threshold (a plan reference, not a peer comparison). Peer comparison across teams is a common wrong turn here — teams have different scopes; sustainable capacity is the load-bearing reference.
• Grain: Team × week, aggregated to a 4-week trailing average to smooth incident-week noise.
• Chart family (Ch 4 preview): Small-multiple line per team with the capacity threshold as a reference line, or a plan-variance bar chart with teams sorted by variance ascending. Either delivers the portfolio read.

BRD #2 — Team Lead (operational decision).

• Audience: A single team lead, daily standup or Monday planning.
• Decision: Which agents to pair, coach, or reassign this week.
• Key question: Which agents on my team are handling above or below the team-median ticket load and resolution time?
• Comparison: Agent-level ticket count and resolution time vs. the team median (a peer comparison inside a small group, which works because the peers are directly comparable).
• Grain: Agent × week, current week only, with the prior week as a reference tick.
• Chart family (Ch 4 preview): A dot-plot of agents ordered by ticket count with the team median as a reference line, alongside a paired dot-plot of resolution time. The lead can spot outliers in either direction in under a minute.

The load-bearing observation. The head of support’s comparison is vertical (team vs. its own capacity); the team lead’s comparison is horizontal (agent vs. peer agents). These are two different chart shapes because they answer two different decision questions. One dashboard cannot deliver both without ambiguating one of them — and once the merge happens, the artifact ships with visuals every audience politely tolerates and none acts on. The two-BRD path is cheaper up front, cheaper to maintain, and defensible in either meeting.

The BRD.

• Audience: Leadership team (CEO, CFO, VP Sales, VP Product), quarterly business review.
• Decision: Whether to expand, hold, or contract next quarter’s marketing budget, and where to shift dollars if holding flat.
• Key question: Which channels are earning their spend and which are not, on a metric leadership already trusts?
• Comparison: Return on marketing investment (ROMI) per channel, ranked, against a company-agreed minimum ROMI threshold (or against last quarter’s ROMI for the same channels — whichever the finance-marketing joint agreement uses).
• Grain: Channel × quarter, blended across campaigns. This is the load-bearing field.
• Time window: Trailing 4 quarters for context; current quarter for the decision.
• Success criterion: Leadership can name at least one budget shift they would make from the chart alone.

The load-bearing field. Grain. If the analyst answers «campaign-level» instead of «channel-level,» the chart will be a top-N campaigns bar or a scatter of hundreds of campaigns — and leadership does not make campaign-level budget decisions, they make channel-level ones. The wrong grain produces a chart that the audience cannot act on, no matter how technically excellent every other choice is. If the analyst answers «monthly» instead of «quarterly,» the chart will show noise at the level leadership does not care about and hide the pattern at the level they do.

Why grain and not audience or decision? Audience and decision are usually the fields the elicitation call gets right, because the requester names them out loud («leadership readout» is right there in the ask). Grain is the field that gets set silently, by the analyst’s reflex to plot at whatever grain the source table happens to be in. It is the field most likely to be filled in wrong and most likely to invalidate every downstream chart choice when it is. Lina’s chapter (Ch 3) is the same failure at the chart-shape level: five channels with three metrics each on a clustered column, when the audience only needed one metric per channel to make the decision. Get the grain right and the chart family, sort order, and encoding fall into place; get it wrong and no amount of visual polish rescues the chart.