Swiggy
Easy
Swiggy food deliveries are facing a lot of delays from the promised time. What could be the reason?
Step 1. Clarifying Questions (scope the drop)
1. Timeframe. Sudden spike on a specific day, or slow drift over weeks
2. Geography. Pan-India, or 2 or 3 cities only
3. Segment. All cuisines and order values, or concentrated in a slice
4. Magnitude. 5-minute average delay or 25 minutes
5. Baseline. Anomaly, or expected behaviour for monsoon, IPL, festival night
If the delay is isolated to 2 cities on weekends, you are solving a very different problem than a 15% all-India shift.
Step 2. Break the order journey into 4 funnel stages
Order placed → Restaurant accepted → Food ready → Rider pickup → Customer delivery
Step 3. Track 3 key metrics at each stage
Stage 1. Order Intake and Restaurant Acceptance
a) Time from order placed to restaurant acceptance (p50 and p90)
b) Restaurant rejection rate
c) Restaurant offline rate in the zone
Stage 2. Kitchen Prep
a) Gap between actual prep time and ETA model prediction
b) p90 prep delay distribution
c) Multi-order queue length per restaurant
Stage 3. Rider Assignment and Pickup
a) Time to assign a rider after food is ready
b) Active rider density in the zone (riders per square km)
c) Rider wait time at restaurant pickup
Stage 4. Last Mile Delivery
a) Rider speed (km per minute) vs baseline
b) Per-stop time on batched orders
c) Distance to customer vs baseline
A single metric is just data. A combination of metrics across stages is a hypothesis. Walk these:
1. Restaurant accept time UP and offline rate UP. Restaurant supply problem. Likely a commission change or restaurant strike.
2. Prep time gap UP and queue length UP at restaurants. Kitchen overload or ETA model has gone stale.
3. Time to assign rider UP and rider density DOWN. Rider supply shortage. Likely attrition, fuel hike, or a competitor running a bonus.
4. Rider wait time at restaurant UP and per-stop time UP. Routing or batching algorithm is pushing too many orders per rider.
5. Last mile speed DOWN and per-stop time UP only in specific zones. External cause. Monsoon, road closure, or local event.
6. ETA error UP across all 4 stages uniformly. ETA model drift after a recent deployment.
7. Delay confined to 1 or 2 cities. Geographic cause. Local festival, regulation, or rider strike.
Accordingly, you can give suggestions
Community Answers (5)
Clarifying Questions: 1. Is it happening in a particular geography? 2. Is it for short distance or long distance deliveries? 3. What is the timeframe of the drop? for one month or one week? Potential Factors: System Related: - Is there a change in how the metric is calculated? - Is the data pipeline broken? - Was there a recent Prod release? Lets do a Funnel Analysis to identify the core issue Estimated Promised Time is calculated from - Prep Time - Travel Time - Pickup & Delivery Time The delay can happen in any of the factors: 1. PREP TIME -Are the restaurants not aware of the prep and packaging policies? If it it was working before, then we can conclude that no change has happened here. 2. TRAVEL TIME- - Is there a metro contruction going in the city so that all deliveries are taking time - Is there a protest in any part of the city? - Is there an extreme weather condition like severe heat or cold or Fog which is not getting captured? 3. PICKUP/ DELIVERY TIME- Apartments and Societies have set up a new process for the delivery executives to enter by asking for approval, which is taking time leading to delay in the promised delivery time
The RCA could be around: 1. Are all swiggy deliveries affected pan India or specific cities only? 2. Is the model accounting for all the variables or is something important not accounted for, in the ETA prediction? 3. Are the delays being experienced by users who are greater than 5-6 km away from the restaurants or are even closer locations affected? 3. Delays since when? Is this seasonal because of the weather? 4. Are these delays affecting those areas where a construction is going on leading to closures and traffic congestion? 5. Have the SOPs changed for the driver operations? 6. There is a shortage of driver staff
Potential reasons: 1. Outdated model which is not accounting for new variables that would affect the ETA 2. Changes in road infrastructure, eg: could be that a few roads are blocked due to metro construction 3. Driver attrition due to which new drivers are still getting used to the delivery model. Even the number of drivers are less 4. Batch processing of orders 5. New restaurants are trending for which the model does not have the correct data for prep time, packing time etc. 6. Some restaurants are becoming more popular and hence have increased influx of orders from all platforms: swiggy/ zomato etc
The root cause could be a Swiggy issue or Rider issue or Restaurant issue or Other (external): Swiggy issue: Wrong ETA estimates - model drift (outdated algorithm) Rider issue: 1. Stacking / Batching of orders 2. Delivery Executive Shortage 3. Greater distance from restaurant 4. Road traffic and infrastructure 5. last mile location and delivery (handover issues) Restaurant Issue: 1. Staffing shortage 2. Inventory Stockouts 3. Dine-in prioritization over delivery (sidelined) 4. Packaging constraints (material shortage) 5. Inefficient handover staging (labelling and biling, etc.) Other Issues (external): Data drift : structural changes such as road closures, permanent rerouting, etc.
Potential Reasons: 1. Promise algorithm malfunctioning 2. Batching of orders leading to delays 3. Any major disruption: Power Outages, Traffic jams, Route Diversions 4. Significant attrition in the delivery fleet. New delivery fleet has a learning curve 5. Change in the mix of last mile delivery vehicles used by the riders