Prompt 14 — Results, analytics, reporting, and imported evidence

Required context

• Thomas-compatible concepts and current imported report flow
• Native question/scale/source model
• Anonymity and data-visibility policy
• Product visualization rules
• User, manager, and HR permissions

Prompt

Build an interactive, evidence-led result experience that modernizes the supplied 360 report concepts without overstating precision or comparability.

1. Results narrative

Organize individual results into:

1. How to read this report: purpose, observation period, scale, source, sample, thresholds
2. Big picture: competency/construct pattern
3. Strengths: consistent high evidence
4. Differences in perspective: self versus others and group variation
5. Role priorities: importance versus observed behavior where collected
6. Development themes: low/high-priority patterns and qualitative evidence
7. Reflection and PDP handoff

Avoid labeling every difference a blind spot. Explain uncertainty and alternate interpretations.

2. Required views

• Competency/construct overview
• Self versus others gap
• Relationship comparison
• Item distribution and Not Observed rate
• Top/bottom or strongest/development items with filter and context
• Importance ranking/importance rating comparison if supported
• Qualitative themes with source counts and evidence snippets
• Longitudinal trend by native family/approved bridge
• Imported evidence timeline
• Accessible data table for every visualization

3. Source and instrument lineage

For every result show:

• Native or external source
• Instrument/report type and version when known
• Import date and original period
• Scale and transformation, if any
• Question/competency mapping confidence
• Whether trend comparison is allowed

Thomas data remains external evidence by default. Preserve original values and source documents. Do not silently normalize it into native trend lines.

4. Anonymity and qualitative safety

• Suppress small groups.
• Guard against difference attacks across filters.
• Redact or moderate identifying/sensitive free text according to policy.
• Show theme counts and uncertainty.
• AI summaries link to source evidence available to the viewer and support correction/reporting.
• Do not expose a hidden respondent through ranges or comments.

5. Interpretation quality

• Show valid N and Not Observed rate.
• Show distribution or range where useful.
• Avoid visual scales that exaggerate 0.1–0.2 differences.
• Distinguish “meeting role expectation” from benchmark claims.
• Do not create a single overall employee or leader score unless explicitly validated and approved.
• Do not rank people or engineering teams.

6. Organization analytics

For HR/authorized roles, provide aggregated:

• Program/campaign participation and coverage
• Construct trends by eligible cohorts
• Not Observed and item-quality signals
• Fatigue/exposure metrics
• PDP activation and follow-up completion
• Source/instrument filters
• Threshold and cohort-size explanations

Do not provide drilldowns that reconstruct individuals.

7. Reports and exports

• Responsive web report
• Accessible print/PDF style
• Authorized CSV/data export
• Branded template by plan
• Source, generated date, period, definitions, and confidentiality labels
• Export audit events
• Offline/static assets bundled and tested; fix the unstyled export problem visible in the current pack

8. Tests

• Score and aggregation calculations
• Not Observed exclusion
• Suppression and difference-attack cases
• Imported lineage and mapping
• Role-specific visibility
• Chart/table equivalence
• PDF/print/offline assets
• Large comments and localization

Acceptance gates

• A subject can understand evidence, uncertainty, and next steps without a facilitator.
• Every score has source, N, period, and observation semantics.
• Imported and native evidence cannot be confused.
• PDF/static export preserves styles and accessibility metadata where supported.