Quality Program — Case Study
Program design

Building a quality
program from zero

Designed and launched a scalable, data-driven quality framework for a team — from blank page to A/B-tested scoring system, structured peer review cadence, and a mentorship program. Delivered across three phases with full documentation and calibration infrastructure.

3
Framework iterations
5
Scoring categories
4
Senior mentors
10
Reviews / agent / month

What I brought to this

Program design A/B testing methodology Rubric development Cross-team collaboration Data-driven decision making Mentorship framework design Documentation Calibration design

What I learned

Ship early, iterate with data

I have high standards and a strong pull toward getting things right before releasing them. This project pushed me to test sooner than felt comfortable — and the data that came back from the first rubric was the only thing that made the second one better. I learned to treat iteration as a feature, not a failure.

Know when to pivot vs. when to fix

The initial qualitative rubric wasn't producing usable data. Rather than spending more time refining the wrong system, I benchmarked against adjacent teams and made the call to move to A/B testing two philosophically different frameworks. Pivoting faster when the signal is clear is something I now do more deliberately.

Trust the research over the room

I consulted peers and adjacent teams, but ultimately the decision about which scoring framework to pursue was grounded in the behavioral psychology behind rating scales — not just in what other teams preferred. Knowing when to trust the research over consensus is a skill I developed through this project.


How it came together

Three distinct phases, each building on what the last one taught. The goal throughout: move from subjective coaching to a transparent, scored, data-driven quality infrastructure.

01
Phase 1
Qualitative 3-point rubric

Built a rubric for peer mentors reviewing new specialists across investigation, communication, documentation, and strategy — scored 1–3 with written notes per ticket.

Learning: qualitative notes don't produce actionable data at scale
02
Phase 2
Cross-team benchmarking

Consulted adjacent quality teams. Identified that existing systems were built for different contexts — and that this framework would need to weight technical depth and documentation differently.

Learning: borrowed frameworks don't always transfer — context matters
03
Phase 3
A/B test: two frameworks in parallel

Designed two competing systems — a BARS (Behaviorally Anchored Rating Scale) and a Binary checklist — tested side-by-side across two cohorts, measuring scoring alignment, grader sentiment, and coaching value.

Finding: BARS surfaces performance gaps the binary system misses

What the data showed

Both rubrics were evaluated for scoring consistency, grader experience, and coaching value. The BARS system was more demanding to apply — and more revealing.

BARS — sliding scale
Growth and nuance model
Surfaces quality gaps hidden by simpler systems
Distinguishes checking boxes from true mastery
Passing threshold: 75%
Harder to apply consistently on simple tickets
Binary — checklist
Safety and compliance model
Faster to grade, easier to calibrate on
Removes grader bias from middle scores
Passing threshold: 86%
Inflates scores — strong and average performers look similar
Key finding: a 35-point gap in "excellent" ratings between the two systems

One specialist's excellent rating dropped from 55% under binary to 20% under BARS — not because their work changed, but because the binary system was rewarding completion rather than quality. BARS revealed who was meeting the standard versus who was exceeding it. That distinction is exactly what a coaching program needs.


The shift this program created

Before
No formal quality process
Coaching was reactive and volume-focused
No scoring, no shared standard, no calibration
New hires had no visibility into how their work would be judged
After
Scored, behaviorally anchored rubric
Monthly review cadence with senior mentors
Auto-calculated scores in a live performance sheet
Transparent criteria shared with the team from day one