Seat Change Optimizer: From Vague Brief to Live Deployment
Led end-to-end design and deployment of an MILP optimizer to reduce gate agent workload at a top US airline.
Athul P M
Senior Associate Consultant at Indus Insights
From their time as
Senior Associate Consultant
Indus Insights • 2025
Overview
Athul was brought onto a project with a single brief: reduce gate agent workload at a top-three US airline. Gate agents were handling last-minute seat changes manually at the gate, a task that was consuming significant time per passenger and creating downstream boarding delays.
The Story
Athul was brought onto a project with a single brief: reduce gate agent workload at a top-three US airline. Gate agents were handling last-minute seat changes manually at the gate, a task that was consuming significant time per passenger and creating downstream boarding delays.
He started by mapping the problem from first principles. Working with operational heads, he ran analytics on time spent per task and isolated seat changes as the primary driver of agent workload. His research into boarding patterns and passenger behavior identified two populations driving the bulk of changes: passengers with special service requests (wheelchairs, infants, pets, language needs) and families traveling with children under 15.
With the problem scoped, Athul designed the solution. He taught himself MILP optimization through NPTEL lectures, connected with the client's internal operations research team, and built an automated seat-change optimizer that could pre-assign optimal seats before the gate opened. He also built a decision interface that showed operational heads which changes were suggested, the reasoning behind each, and the factors to accept or override.
Stakeholder Alignment
The harder challenge was getting operational heads to accept the optimizer's logic. The algorithm produced near-optimal results, not perfect ones, and stakeholders initially pushed back, expecting every preference to be honored. Athul translated the mathematics into plain business terms, explaining the trade-off between optimality and perfection and securing alignment on what the non-negotiables actually were.
He coordinated across operations, data science, and engineering teams, each with their own perspective on the solution, and kept the project moving by moderating competing views toward a shared outcome.
A/B Test and Deployment
To isolate the optimizer's impact, Athul designed a rigorous A/B test across 4,200+ flights at ORD Airport over 10 days. He normalized for flight body type (wide-body, narrow-body, regional jet), haul length, SSR population distribution, and family composition, then validated the two groups were statistically comparable using a t-test before proceeding.
The test showed a 14% net reduction in gate-side seat changes for the target populations. The optimizer is now in live deployment. Athul also built a Power BI dashboard quantifying impact across 10+ KPIs in real time for the operations team.