Rewarding the teachers
who make students stay
Teacher quality was invisible to the business
No attribution model
Renewals were tracked at the business level — never tied to individual teachers. A teacher with 90% renewal rate earned the same as one with 40%.
Misaligned incentives
Compensation was based on hours taught, not outcomes delivered. High performers had no reason to go further. Low performers faced no consequence.
Gaming risk
Any reward system could be gamed. Teachers could pressure students for ratings or selectively teach high-renewal cohorts. The model needed to be robust by design.
Fairness across cohorts
A teacher with 10 students can't generate the same referral count as one with 50. Criteria needed to normalise for class size and tenure to be perceived as fair.
Cross-functional from day one
Adnan Murad
System design, criteria logic, teacher dashboard UX, stakeholder alignment
Mehul Shah
Roadmap ownership, business case, feature prioritization across 18 months
Priya Reddy
Criteria modeling, metric definitions, threshold calibration and validation
Karan Bose
Teacher communication, rollout strategy, edge case handling and escalations
Transparency as a design principle
Four dimensions, one composite score
Recognition is a system, not a sticker.
Attendance, pace, effort, punctuality and scores were already tracked — none of it came back to the student. The badge system turns data the platform already had into something a kid can see, chase and share, and something a parent and teacher can celebrate with them.
“Five badges. One monthly goal. A next step within reach for every kid.”
What the program taught me
Every criteria decision had to pass a fairness test across teacher cohort sizes, subjects, and tenures. We learned to prototype with data before committing to any threshold.
The single highest-impact decision was making scores explainable at the student level. Teachers went from suspicious to advocates — because they could see inside the system.
Every time a criterion was gamed, it surfaced something real that was missing from the model. Rather than fighting gaming, we used it as a signal to improve the criteria themselves.
A system right in August 2023 was wrong by January 2025. Teacher populations shift, business priorities evolve. Incentive systems need owners — not just launchers.