Teardown: ride-hailing and food delivery
A ride or a meal arriving looks like one app. It is really three apps, one for the customer, one for the driver or rider and one for the merchant, plus a control room, held together by four mechanisms: the ETA promise, dispatch, pricing and ratings. This lesson tears down Uber, Grab, Pathao and foodpanda with the method from pm-40, and ends with what Mangrove's Rider app should copy.
By the end you can
- Map the customer, driver or rider, and merchant apps, and the job each one serves
- Break an ETA into its parts and say which side controls each part
- Explain why the nearest driver is often the wrong one, and when batching two orders pays
- Explain surge pricing as a feedback loop, and choose other levers when raising prices would hurt
- Judge what ratings, incentives and cash do to the supply side, and write copy, adapt and avoid for a delivery app
The everyday picture
A busy restaurant on a Friday night has a host at the door, a kitchen and a team of waiters. A poor host seats each new table with whichever waiter is standing closest. A good host looks ahead: this waiter will be free in two minutes, that kitchen station is backed up, so seat the big table over there. Waiters who stand around waiting for food cost money; customers who wait for a waiter leave unhappy.
When a storm makes everyone come in at once, the restaurant has a few levers. It can raise prices, call in more staff with extra pay, close the waiting list, or tell new guests that the real wait is 40 minutes. Ride-hailing and food-delivery apps are that host, kitchen pass and waiting list, run by software across a whole city, thousands of times a minute.
Three apps and a control room
Decision question first. Imran's: "On-time 'Now' deliveries are at 88% this month. What should the Rider app and the Store app copy from ride-hailing and food delivery to reach 92%?" Mangrove has about 600 riders and promises 30-minute "Now" delivery.
Then the sides. The customer wants a price and an arrival time she can trust. The driver or rider wants steady earnings per hour and as little unpaid waiting as possible. The merchant (a restaurant or shop) wants orders it can prepare without chaos at the counter. The platform's operations team wants supply and demand balanced in every neighbourhood. Each side has its own app: Uber's rider and driver apps, a merchant app or tablet in each restaurant, and dashboards for the control room. The hard design work lives on the supply side.
Here are the four companies, with the facts we checked.
- Uber completed about 3.5 billion trips in the third quarter of 2025, for 189 million monthly users. Its gross bookings were about $49.7 billion that quarter, about half from rides and half from delivery. Dividing its reported revenue by gross bookings gives about 31% for rides and 19% for delivery. That is a rough measure: accounting rules differ by country, so it is not the commission a driver sees. The Uber system design teardown covers how matching works inside.
- Grab started as MyTeksi, a taxi-booking app, in Kuala Lumpur in June 2012. It tested GrabBike motorcycle rides in Ho Chi Minh City in November 2014, launched the GrabPay wallet in November 2017, took over Uber's Southeast Asian business in March 2018, and launched GrabFood in May 2018. It now operates in eight countries, with payments, lending and a digital bank alongside rides and food. Its 2024 revenue was about $2.8 billion, with a net loss of about $158 million.
- Pathao was founded in Dhaka in March 2015 as a delivery service using motorcycles and bicycles for online shops. It added motorcycle ride-sharing in mid-2016, before the rules allowed it. The transport authority first called ride-sharing illegal; guidelines followed in December 2017, and it received its licence in February 2018. By March 2018 it said it had more than 100,000 drivers and about a million users. Its food delivery started in January 2018, using bicycles to beat traffic. Its own wallet launched in 2018 so customers could pay riders in the app, but the central bank reportedly refused permission for a closed wallet, so cash stayed central.
- foodpanda was founded in Singapore in 2012 and bought by Delivery Hero in 2016. In a 2022 case study it described working in more than 300 cities in 12 countries with over 115,000 restaurants and over 80,000 riders. In May 2024 Uber agreed to buy its Taiwan business for $950 million; Taiwan's competition regulator blocked the deal that December.
Two patterns stand out. Every one of them started with one job and added others: Pathao went from parcels to rides to food, and Grab from taxis to a wallet, food and a bank. And every one adapted to local conditions: motorcycles where traffic is dense, bicycles for food, cash where wallets were not allowed or not trusted.
Quick check
Why does a teardown of a delivery app need to study the rider and merchant apps, not only the customer app?
The customer sees a promise; the merchant and rider make it come true or break it. Preparation, assignment and the ride are where minutes are won and lost, so that is where the mechanisms are. Review counts and who pays more don't tell you where the design problems are.