Restaurant capacity
Peak-hour order volume may exceed the restaurant's practical preparation capacity.
PRODUCT DISCOVERY · DATA ANALYSIS · EXPERIMENT DESIGN
Swiggy Delay Recovery
A product and analytics case study for reducing uncertainty before a late order becomes a trust failure.
Independent portfolio concept using synthetic data. Not affiliated with or endorsed by Swiggy.
THE CUSTOMER PROBLEM
The worst part is not always waiting. It is finding out too late.
A customer may willingly wait for the meal they actually want. Trust breaks when the promised time becomes unreliable and an eventual cancellation arrives after the customer has lost both time and alternatives.
Product insight Give customers an honest estimate and an early choice—before support becomes the only path.
HYPOTHESES
Start with the preparation signal, then inspect the full journey.
Preparation estimates
Restaurants may enter estimates that do not adjust enough for order size and live demand.
Promise logic
A single delivery estimate can miss peak periods, large baskets and compounding journey delays.
SYNTHETIC SCENARIO MODEL
A transparent simulation for testing the analysis workflow.
I constructed 600 orders so peak periods and larger orders carry more delay and cancellation risk. The outputs show how I would segment and compare the journey; they do not discover or prove relationships at Swiggy.
These are scenario outputs created from explicit assumptions—not Swiggy operational findings or measured company performance.
PROPOSED MVP
Make the changing promise visible—and make recovery easy.
Scenario-aware estimate
Use a transparent rule-based estimate in the analysis; a future product could evaluate live demand, order size and restaurant signals.
Early warning
Explain the likely delay as soon as confidence drops, not after the promise is missed.
Customer choice
Let the customer continue waiting or cancel without needing to contact support.
Fast recovery
Show the available refund paths clearly so the customer can order again.
PILOT DESIGN
Test the recovery journey with limited exposure.
Randomize eligible at-risk orders between the current journey and the new recovery experience. Begin with a 10% exposure cap for operational safety, then calculate the required sample from eligible volume, baseline reorder rate and the minimum meaningful lift.
PRIMARY SUCCESS METRIC30-day reorder
Measure lift among customers affected by a restaurant- or platform-initiated cancellation. Set the minimum detectable effect after baseline and power analysis.
Guardrails
- No rise in avoidable cancellations
- ETA error and promise accuracy remain acceptable
- Support-contact and post-recovery satisfaction
- Refund, compensation and contribution-margin cost
- No harmful repeat exposure within the test window
LEADING INDICATORS
Know whether recovery is working before waiting 30 days.
Warning timing
Share of at-risk customers warned before the original promise is missed.
Self-serve recovery
Recovery completed without support, plus time to a clear wait-or-cancel decision.
Ability to reorder
Time until funds are available, replacement-order rate and time to replacement.
METHODOLOGY NOTE
Transparent assumptions make the work stronger.
The scenario assigns higher support contact and lower satisfaction and reorder behaviour to cancelled orders so the comparison workflow can be demonstrated. It does not establish causality or represent access to Swiggy systems, internal metrics or measured impact.
CURRENT EVIDENCE
One clear, recruiter-safe source of truth.
This page reflects the credibility review completed on 10 August 2026. The earlier downloadable deck is withheld until its modeled findings, estimate label and pilot measurement language are revised to the same standard.