Skip to content
Case study

Swiggy: the order that's always 5 minutes away

Diagnosing late deliveries when the ETA keeps lying.

Decomposes 'delivery is late' into the four handoffs where time leaks, and argues that managing the ETA promise is sometimes a better fix than shaving the minutes.

Role
Product teardown — operations & RCA
Timeline
RCA exercise
Year
2025
Domain
Food delivery · Operations
Funnel decompositionRCAExpectation management

The complaint

'It's late' is four problems wearing one trench coat.

Delivery time isn't a number; it's a chain — order acceptance, restaurant prep, rider assignment and pickup, and the ride to the door. 'Orders are late' is meaningless until you know which link is leaking time. The same complaint can be a slow kitchen, a rider-supply gap, or a routing miss, and each has a completely different fix.

Decompose the clock

The four places time leaks

  • Restaurant prep: the most variable and least controllable link. Slow or under-staffed kitchens, especially at peak, dominate the delay.
  • Rider assignment: the gap between food-ready and a rider in hand — a supply/density and batching problem.
  • Pickup wait: rider arrives, food isn't ready (or vice versa) — a synchronization problem between two parties who don't talk.
  • Last mile: the actual ride, mostly a routing and traffic problem and the easiest to estimate.

Tradeoffs

The non-obvious move

Chose

Sometimes fix the promise, not the time

over Always trying to deliver faster

A 35-minute delivery promised as 25 feels late; promised as 40 feels early. When the minutes are structurally hard to shave (a slow kitchen at peak), an honest, slightly padded ETA buys more satisfaction than a heroic, often-broken one. Expectation is a product surface.

Chose

Attack prep-time variance at specific restaurants

over A blanket 'speed up delivery' push

The delay is concentrated, not uniform. A handful of high-volume, high-variance kitchens cause a disproportionate share of late orders — fixing their prep flow (or their displayed prep time) beats optimizing the whole fleet.

Reflection

The counterintuitive product call

The instinct is to make everything faster. The sharper move is to recognize that a trustworthy ETA can beat a faster, less reliable one — and that the link to fix is rarely the one the user is yelling about. [PLACEHOLDER: add a real Swiggy data point or personal anecdote if available.]

Next · build

Noor

An AI copilot that drafts the reply before you've finished reading the thread.

Continue