Domains 1 and 2: Audience and Context Assessment (Locked)
Covers Domain 1 (Audience Adaptation) and Domain 2 (Message Alignment). Both use the shared audience taxonomy and message-clarity anchors below.
Audience Taxonomy (Shared with Viz Planner)
| Audience | Time Budget | Filters | Charts | Interaction Style |
|---|---|---|---|---|
| Executive | 5-30 seconds | 0-2 | 4-6 | Minimal, hover tooltips only |
| Manager | 1-3 minutes | 2-4 | 5-8 | Moderate; filter actions, drill |
| Analyst | 3-15 minutes | 4-8 | 6-12 | Heavy; parameter actions, set actions, search |
| Operations | 5-30 seconds | 1-2 | 3-5 | Alert-driven; status color + threshold only |
| External / Public | 30-90 seconds | 0-1 | 3-5 | Minimal, pre-scoped, storytelling |
Audience is either stated by the user or inferred from dashboard signals (title language, chart count, filter count, complexity).
Domain 1: Audience Adaptation
What it measures. How well the dashboard fits its stated or inferred audience per the taxonomy above.
Signals evaluated (image + optional XML):
- Chart count vs. audience chart budget.
- Filter count vs. audience filter budget.
- Complexity level (calc-heavy, dual-axis, small-multiples) vs. audience skill level.
- Reading-time estimate vs. audience time budget.
- Layout scan path (visual hierarchy) vs. audience decision-making style.
Score anchors:
| Score | Description |
|---|---|
| 10 | Dashboard fits the audience precisely. Chart count, filter count, and complexity all inside the audience's window. |
| 8-9 | Strong fit with one minor mismatch (e.g., one extra filter for Executive; one complex calc for Manager). |
| 6-7 | Solid fit; one clear mismatch that increases reading time by ~30%. |
| 4-5 | Developing; multiple mismatches. Audience will need effort the design does not support. |
| 2-3 | Weak; audience mismatch limits usability significantly. |
| 1 | Fundamental audience miss (e.g., Analyst-tier scatter plot on an Executive dashboard). |
Common downgrades:
- Executive dashboard with 8+ charts or a scatter/histogram/correlation view.
- Analyst dashboard with only KPI tiles and no drill capability.
- Operations dashboard without a clear status-color + threshold pattern.
- Multi-page navigation on an Executive dashboard.
HC#9 partial-info handling. If audience is stated, honor it. If stated audience contradicts observable design (Executive with 20 charts), flag the conflict in the Domain 1 rationale rather than lowering the score silently.
Domain 2: Message Alignment
What it measures. How clearly the dashboard communicates its core message or decision-support intent.
Signals evaluated:
- Title clarity. Does the title name a decision, question, or metric?
- Subtitle or caption. Does it disclose the audience, time window, or filter context?
- Primary-metric hierarchy. Is the most important number the largest, first, or most-emphasized element?
- Supporting-chart alignment. Do secondary charts elaborate on the primary metric, or do they compete for attention?
- Empty-canvas discipline. Are unused zones minimized? Are decorative elements minimized?
Score anchors:
| Score | Description |
|---|---|
| 10 | Message reads in under 5 seconds; every element supports the primary decision or question. |
| 8-9 | Message reads in under 10 seconds; one or two elements are decorative or off-message. |
| 6-7 | Message reads with modest effort; 3-4 elements are off-message or ambiguous. |
| 4-5 | Message unclear on first pass; requires scanning to identify the primary metric. |
| 2-3 | Message competes with too many elements or is buried. |
| 1 | No discernible message; dashboard reads as a data dump. |
Common downgrades:
- Title is a field name ("Sales Dashboard") rather than a decision or question.
- 4+ competing primary metrics at similar visual weight.
- Decorative elements (logos, banners) larger than the primary KPI.
- Missing subtitle when audience or time-window matters.
- Empty zones with placeholder text or unused white space that suggests the dashboard is incomplete.
Audience Inference Signals (When Not Stated)
If the user did not state the audience, infer from these signals in the image, in order of weight:
- Chart count. 4-6 → Executive or External. 5-8 → Manager. 6-12 → Analyst. 3-5 → Operations or External.
- Filter count. 0-1 → Executive or External. 2-4 → Manager. 4-8 → Analyst.
- Title language. "Overview" or "Summary" → Executive. "Detail" or "Deep Dive" → Analyst. "Status" or "Monitor" → Operations.
- Interaction affordances visible. Multiple filter controls + parameter dropdowns → Analyst. Minimal or none → Executive.
- Color palette. Alert-heavy (orange, magenta) with shape encoding → Operations.
Confidence flag when inference is Low (contradicting signals). Ask the user for audience if truly ambiguous per HC#9.
Calibration Examples
Real-world signal-to-score mappings for both domains. Use as anchors when the current image matches or approximates one of these shapes.
Domain 1 Calibration
Example 1.1 → Score 9. Executive dashboard, image shows 4 KPI tiles on the top row, 2 supporting trend charts below, one filter for time period, no additional filters visible. Signal: chart count 6, filter count 1, complexity minimal. All three inside Executive time budget. Strong fit.
Example 1.2 → Score 6. Manager dashboard stated by user, image shows 9 charts including one scatter plot and one dense heatmap. Signal: chart count exceeds Manager budget by 1, and analytical-tier chart types dilute the 1-3 minute time window. One clear mismatch.
Example 1.3 → Score 3. Executive dashboard stated by user, image shows 14 charts including three scatter plots and a 30-row detail table. Signal: chart count 2.5x Executive budget, three analytical-tier charts require Analyst time budget, detail table is a Manager or Analyst artifact. Fundamental audience miss.
Domain 2 Calibration
Example 2.1 → Score 9. Title reads "Q4 Revenue Growth: On Track for 8% YoY." Primary KPI (Revenue) is the largest element, centered top-left. Three supporting charts elaborate (regional breakdown, product-category breakdown, month-over-month trend). No decorative elements larger than a chart title. Signal: title names the decision, primary KPI anchors hierarchy, secondary charts elaborate. Message reads in under 5 seconds.
Example 2.2 → Score 6. Title reads "Sales Dashboard." Two primary KPIs (Sales and Profit) at similar visual weight. Company logo is 2x the size of the primary KPIs. Signal: title is a field name, two competing primary metrics, decorative element dominates. Message unclear on first pass.
Example 2.3 → Score 2. Title is "Overview." Four primary metrics tied at the same size. No subtitle. Empty zones scattered across the layout with placeholder text ("chart here"). Signal: no discernible primary metric; layout reads as incomplete dashboard.
Ask-For-Audience Cadence
Skill asks for audience only when inference confidence is Low AND the domain-1 or domain-2 score would materially change based on audience. Example ask:
"Who's the primary audience for this dashboard: Executive, Manager, Analyst, Operations, or External? I can score without it, but audience-fit accuracy improves with the answer."
Do not ask for audience on runs where confidence is High or Medium; the inference is good enough.
Refuse to Fabricate
Never score Domain 1 or 2 on general impression. Cite specific image signals. Flag confidence when inference is Low. Do not invent an audience-mismatch to justify a lower score, and do not paper over a genuine mismatch to soften the critique.