About This Project
This analysis was developed independently to evaluate a proposed redesign of peer feedback methodology during the performance cycle. The existing format—a 5-point Likert agree/disagree scale—was producing heavily skewed data that offered little diagnostic value for development conversations or L&D resource targeting.
I designed and ran an A/B pilot with 169 participants comparing the existing scale against a dual-axis Frequency × Impact model, then analyzed results across overall averages, role type, location, and job family to identify where the new format produced meaningfully different signals.
This demonstrates: experimental learning design thinking, quantitative data analysis, iterative program development, and cross-functional insight synthesis. It was never formally a “project request”—I identified the problem, proposed the methodology, and ran the test.
The Problem
The existing peer feedback format used a 5-point agree/disagree Likert scale to evaluate core company values. The format produced limited diagnostic value:
| Rating | Usage |
|---|---|
| Strongly Agree | 85%+ |
| Somewhat Agree | ~13% |
| Neutral / Disagree | ~2% |
Ceiling concentration confirmed: 98%+ of ratings landed positive, making it impossible to distinguish high performers from adequate performers or to target L&D resources meaningfully.
The Hypothesis
A dual-axis Frequency × Impact rating scale (Option B) would produce more nuanced, actionable feedback data—enabling richer development conversations and a foundation for learning resource targeting.
A/B Test Design — 169 Participants
| Option A (Current) | Option B (Future State) | |
|---|---|---|
| Scale | Strongly Agree → Strongly Disagree (5-pt) | Frequency (4-pt) + Impact (4-pt), separately |
| n | 89 participants | 80 participants |
| Frequency labels | — | Consistently / Frequently / Sometimes / Rarely |
| Impact labels | — | Exceptional / Effective / Developing / Limited |
Option B: Results
Option B surfaced meaningful differentiation across values, job families, roles, and locations—data that could directly inform where to focus L&D resources. Scores varied by competency area, with some showing notable gaps between frequency of behavior and perceived impact—a pattern the original format made invisible.
Analysis Framework: Frequency × Impact Matrix
| Quadrant | Pattern | Recommended Intervention |
|---|---|---|
| High Freq / High Impact | Consistently performing with strong outcomes | Leverage strengths; peer upskilling; explore leadership |
| High Freq / Low Impact | Behavior is frequent, but outcome is weak | Skill refinement—how to perform the behavior more effectively |
| Low Freq / High Impact | Has capability; not applying it consistently | Habit building—perform behavior more regularly |
| Low Freq / Low Impact | Rarely performing, limited impact | Foundational skill-building, coaching, targeted feedback |
Open-Text Themes — 100 Responses Across A & B
Key finding: Strongest responses used specific situations with measurable impact. Weakest were generic praise with no actionable detail—reinforcing the need for structured behavioral frameworks in feedback prompts.
What I Built Beyond the Pilot
- Individual reporting template—personalized H1/H2 Frequency and Impact averages by competency area for each team member
- People manager support framework—guidance on using results to lead development conversations
- Resource-mapping framework—competency-specific learning resources mapped to Frequency and Impact score combinations
This Demonstrates
- Experimental learning design—identified a real problem, proposed a testable hypothesis, ran a controlled pilot
- Quantitative data analysis—multi-variable breakdowns across job family, location, role, and competency
- Instructional systems thinking—output was a scoring model with a full intervention matrix, not just a report
- End-to-end ownership—from problem identification through pilot design, execution, analysis, and recommendation
- Proactive initiative—built and ran without a formal project request