name: sf-people-analytics-quality-review
description: Use when you need to check whether reports are trusted, performant, governed, and safe to use for decisions.
version: 1.0.0
author: SahirVhora
license: MIT
metadata:
hermes:
tags: [sapsf, analytics, agent-skill, consulting]
related_skills: [sf-config-debt-workshop]
People Analytics Quality Review
Reviews People Analytics, Report Center, and story reports for duplicate reports, stale owners, missing data definitions, weak filters, row-level security gaps, slow queries, and inconsistent KPI logic. Produces a report rationalisation plan and governance model so HR leaders know which dashboards to trust.
When to Use
- Check whether reports are trusted, performant, governed, and safe to use for decisions.
- Client asks for an evidence-backed review in the Analytics area
- Preparing a workshop, release gate, audit pack, or remediation plan
- Converting raw SF configuration into a client-safe recommendation
Prerequisites
- Inputs: Report catalog, story definitions, user access, refresh schedules, business KPIs
- Expected outputs: Report health score, duplicate report map, data trust issues, governance actions
- Confirm client audience and whether output should be board-level, technical, or mixed
- Never store credentials or employee-sensitive data in the repo or final deliverable
Workflow
- Catalog - gather evidence, classify impact, and create a client-safe output.
- Trust Check - gather evidence, classify impact, and create a client-safe output.
- Performance - gather evidence, classify impact, and create a client-safe output.
- Governance - gather evidence, classify impact, and create a client-safe output.
Analysis Checklist
- Confirm the configuration objects and source tenant/snapshot date
- Separate configuration evidence from assumptions
- Score findings by business impact, not just technical severity
- Group repeated findings into themes so the client gets a short action list
- Flag internal-only notes before writing the client-facing summary
- Produce remediation actions with owner, effort, dependency, and success metric
Edge Cases
- Two reports use same title but different filters: validate explicitly before final recommendation
- Owner left the business: validate explicitly before final recommendation
- Report includes terminated employees unexpectedly: validate explicitly before final recommendation
- Country filter missing for works council region: validate explicitly before final recommendation
- Story dashboard joins incompatible datasets: validate explicitly before final recommendation
- Scheduled report sends sensitive data to large group: validate explicitly before final recommendation
Example Prompt
Review our People Analytics reports and tell us which dashboards are safe, duplicated, or misleading.
Example Output Shape
184 reports reviewed. 39 duplicates, 22 stale owners, 14 reports with sensitive fields sent by schedule, 9 KPI conflicts. Recommendation: retire 51 reports, certify 23 executive dashboards, and add report owner review every quarter.
Common Pitfalls
- Delivering raw technical noise: Summarise by business impact and put raw details in an appendix.
- Ignoring country or legal-entity variation: Many SF issues are only defects in one population.
- Missing downstream impact: Always map the finding to payroll, compliance, reporting, integration, or user experience.
- No rollback plan: Every remediation step needs a safe fallback.
- No validation step: Re-run the relevant check after fixing config and compare before/after evidence.
Verification Checklist
- [ ] Source evidence captured with tenant/snapshot date
- [ ] Findings scored by severity and business impact
- [ ] Edge cases reviewed explicitly
- [ ] Remediation actions include owner, effort, dependency, and success metric
- [ ] Client-safe summary produced
- [ ] Internal-only notes separated
- [ ] Follow-up validation plan included