Pulse Survey Analysis

Inputs

Company

YYYY-MM

EmpID

Department

Manager

Q1_Manager_Leadership

Q20_Engagement_Referral

Manager

People Leader

Outputs (beside CSV, unless

--out-dir

)

1. [Team] Pulse Survey Report - [YYYY-MM].pdf

   — 8-section team report

2. [Team] Pulse Survey Charts - [YYYY-MM].png

   — 7-panel composite

3. individual_reports/{Manager} - Pulse Survey Personal Report.pdf

   (+ chart PNG)

4. [Team] Pulse Survey Metrics - [YYYY-MM].json

   — Python numbers for

   html-executive-deck-generation

   (always written; override path with

   --json-out

   )

Workflow

Progress:
- [ ] Confirm team name + period
- [ ] Ensure venv deps installed
- [ ] Run pulse_report_runner.py
- [ ] Reply with output paths + 3 one-line findings

1. Setup (once per machine)

~/.cursor/skills/pulse-survey-analysis

python3 -m venv .venv
source .venv/bin/activate
pip install -r requirements.txt

2. Generate report (preferred — matches PDF template)

.venv/bin/python scripts/pulse_report_runner.py \
  "/absolute/path/to/engagement_survey_raw_data.csv" \
  --team "Company" \
  --period "2026-05" \
  --json-out "/absolute/path/to/out/metrics.json"

• --out-dir DIR

  — write outputs elsewhere

• --no-individual

  — team PDF + charts only

• --no-pdf

  — skip PDFs; still write PNG + JSON (faster deck path)

• --json-out PATH

  — metrics JSON for the executive deck (default: beside other artefacts)

Desktop - km’s MacBook Pro

pathlib

'

3. Reply to user

Hard rules

1. Direct Leader score = mean Q1–Q10 only → rank managers by this.
2. Overall average = mean of all Q columns → context only; never sole judgment of a manager.
3. Ratings ≤ 2 are extreme — always surface counts.
4. Q11–Q13 and Q14–Q17 are systemic — do not blame one manager.
5. If manager n ≤ 5, label conclusions “directional only”.
6. Prefer the bundled runner for PDF/PNG fidelity; do not freehand a divergent layout unless the user asks for a different format.
7. When this skill is used to feed

   html-executive-deck-generation

   , the deck may cite only the JSON this runner writes. Do not re-average Q columns in the LLM.

Manual analysis fallback