On this page · 15 sections
- Who this is for
- The three apps and the console nobody quotes
- Meter one: what the map actually costs
- Meter two: the fare engine has a legal ceiling
- Meter three: paying the fleet is a regulated flow
- Meter four: the gig welfare layer, and why it is a product requirement
- What the market rewards
- How we build
- Engagement models
- The stack
- Why eCorpIT
- Related work
- How eCorpIT can help
- FAQ
- References
Summary. An on-demand build in India is priced by four meters that most quotes leave out. The map bill: Google Maps Platform charges $5.00 per 1,000 Compute Routes Essentials calls and $2.83 per 1,000 Places autocomplete requests once the free caps are used, on a price list last updated 7 August 2026. The fare engine: the Motor Vehicles Aggregator Guidelines 2025 from the Ministry of Road Transport and Highways cap dynamic pricing at two times the base fare and require a driver who brings his own vehicle to receive at least 80% of the fare. The payout rail: the Reserve Bank of India's Payment Aggregator Directions of 15 September 2025 put merchant money in an escrow account at a scheduled commercial bank and state that "A PA business shall not carry out marketplace business." And from 13 February 2026, Karnataka charges a 1% welfare fee on every payout an aggregator makes to a gig worker. eCorpIT builds on-demand and marketplace platforms from Gurugram, and this page sets out what the work involves and what it costs to run once it is live.
Who this is for
You are commissioning a marketplace where supply moves: rides, deliveries, home services, field technicians, equipment hire. Someone requests, someone is matched, someone travels, someone gets paid. The demand side is the easy half.
If your model is a directory with a phone number attached, an on-demand platform is more machinery than you need. We will say so on the call.
The three apps and the console nobody quotes
An on-demand product is four applications, and buyers are usually quoted for one and a half of them.
The customer app is the one in the demo. The provider app is harder: it runs for eight hours in a pocket on a mid-range Android phone, on patchy mobile data, holding a GPS lock and a websocket, and it decides whether your unit economics work. The operations console is the one that gets cut from version one and then rebuilt in a panic in month four, because until you have live order state, manual reassignment, a refund path and a fraud queue, your operations team runs the business on WhatsApp.
The fourth is the dispatch service itself, and it is not a screen. It is a matching loop that reads live provider locations, filters by distance and acceptance history, offers the job, handles the decline, and reoffers within a few seconds. Swiggy reported 648,000 average monthly transacting delivery partners in the quarter ended 30 June 2026 and 1,089 orders per dark store per day across 1,171 dark stores in 131 cities. Matching at that density is a systems problem, not a feature.
Budget the console at the same weight as the customer app. That single decision separates the builds that scale from the builds that get rewritten.
Meter one: what the map actually costs
Location is metered per call, and the meter runs forever. Google Maps Platform bills per SKU with a monthly free allowance for each, on rates last updated 7 August 2026.
| Maps Platform SKU | Free events per month | Rate per 1,000 above the free cap |
|---|---|---|
| Places autocomplete requests (Essentials) | 10,000 | $2.83 |
| Geocoding (Essentials) | 10,000 | $5.00 |
| Routes: Compute Routes Essentials | 10,000 | $5.00 |
| Routes: Compute Route Matrix Essentials | 10,000 | $5.00, billed per element |
| Dynamic Maps load (Essentials) | 10,000 | $7.00 |
| Routes: Compute Routes Pro | 5,000 | $10.00 |
| Navigation SDK navigation request | 1,000 | $25.00 |
Rates step down with volume: Compute Routes Essentials falls to $4.00 per 1,000 above 100,000 monthly events and $3.00 above 500,000.
Here is a worked example at stated assumptions, for one city doing 30,000 completed trips a month. Two autocomplete requests, one geocode, three route computations and two dynamic map loads per trip, plus a five-provider dispatch shortlist that bills as five route matrix elements. That comes to $141.50 for autocomplete, $100.00 for geocoding, $400.00 for routes, $650.00 for the matrix and $350.00 for map loads: $1,641.50 a month, about $0.055 per completed trip.
Now turn on in-app turn-by-turn navigation. The Navigation SDK bills 30,000 navigation requests against a 1,000-event free cap at $25.00 per 1,000, adding $725.00 and taking the same month to $2,366.50. One product decision raised the map bill by 44%.
The engineering response is caching and call discipline: cache geocodes against a stable address ID, use session-based autocomplete, shrink the dispatch shortlist, and do not recompute a route on every location ping. This is why the matrix call, not the map on screen, is usually the largest line.
Meter two: the fare engine has a legal ceiling
Surge pricing in India is not a product decision. The Motor Vehicles Aggregator Guidelines 2025, issued by the Ministry of Road Transport and Highways, set the boundaries, and States were directed to adopt them within three months of issue.
| Guideline | The rule | What it means in the code |
|---|---|---|
| Dynamic pricing, clause 17.3 | "a maximum dynamic pricing of two times the base fare" | Hard cap in the pricing service, not a config a growth team can raise |
| Fare floor, clause 17.3 | Minimum 50% below base fare | Discounting has a floor as well as a ceiling |
| Dead mileage, clause 17.2 | Base fare chargeable for a minimum of 3 kilometres | Short-trip fares are computed differently from distance |
| Driver share, clause 17.4 | At least 80% of the fare to a driver using his own vehicle | The take rate is a legal constant, not a lever |
| Fleet vehicles, clause 17.5 | At least 60% where the aggregator owns the vehicle | Two payout models, two ledgers |
| Cancellations, clause 18 | 10% of fare, capped at ₹100, on either side | Cancellation logic needs a published reason code list |
| Licensing, clause 5 and 6 | ₹5,00,000 licence fee, ₹25,000 renewal, security deposit from ₹10,00,000 | A capital line, not a software line |
Two consequences follow for the build. First, the pricing engine needs a policy layer per State, because States adopt the guidelines with their own additions and fare determinations. Second, every fare needs an auditable breakdown stored at the time of the trip: base, distance, time, multiplier, driver share, aggregator apportionment. When a regulator or a driver asks how a number was produced eleven months later, a recomputation is not an answer.
Meter three: paying the fleet is a regulated flow
Moving money to thousands of small earners is where in-house builds most often stall. The Reserve Bank of India consolidated the rules in the Master Direction on Regulation of Payment Aggregators dated 15 September 2025, which repealed the 2020 guidelines and the separate cross-border circular.
What matters to a marketplace:
A non-bank payment aggregator must hold customer money in a separate escrow account with a scheduled commercial bank in India, and the escrow is a designated payment system under section 23A of the Payment and Settlement Systems Act 2007. Nodal accounts are gone.
The Directions define a marketplace as an e-commerce entity providing a platform to connect buyers and sellers, and include a marketplace within the definition of a merchant. They also state that "A PA business shall not carry out marketplace business." Read together, the default architecture is that you are a merchant of a licensed payment aggregator, not an aggregator yourself. Going the other way means an Indian company, an application through RBI's Pravaah portal, a minimum net worth of ₹15 crore at application, and ₹25 crore by the end of the third financial year after authorisation.
Paying an underlying seller out of escrow is permitted on the merchant's specific direction where that merchant has annual turnover above ₹40 lakh, or export turnover above ₹5 lakh. Escrow cannot be operated for cash-on-delivery transactions, which matters if your category still runs cash.
Settlement timing changed and most vendor content has not caught up. The T+1 style timelines everyone quotes came from the 2020 guidelines. The 2025 Directions replace them with the agreement between the aggregator and the merchant, which "should be fair, equitable and must transparently mention the settlement timelines." Instant payout to drivers is now a commercial negotiation with your provider, and it belongs in the contract before it goes in the app. If your platform also runs subscriptions or wallet auto-top-ups, the separate RBI e-mandate engineering checklist covers the authentication and pre-debit notice rules.
Meter four: the gig welfare layer, and why it is a product requirement
Four States have now legislated on platform workers, and the compliance work lands in your codebase rather than in a policy folder.
| State | The levy | Status as of August 2026 |
|---|---|---|
| Karnataka | 1% to 5% of each payout to a gig worker, notified at 1% | Collecting from 13 February 2026 |
| Jharkhand | 1% to 2% of the aggregator's annual turnover, quarterly | Act issued 6 January 2026, start date to be notified |
| Telangana | 1% to 2% on transactions | Bill passed 30 March 2026, Rules not notified |
| Rajasthan | Rate to be notified by the State | Act of 2023, Rules never framed |
Karnataka is the one already running, and its Rules of 19 November 2025 read like a technical specification. Aggregators self-declare the fee within five working days of each quarter end and remit quarterly, with 12% simple interest on late payment. Every payout and its corresponding fee is uploaded to the State's verification system within seven days of the transaction, with monthly reconciliation and a quarterly return. Registration was due within 45 days of commencement, and the worker database with it.
Two Rules change the product. Termination of a worker needs 14 days' prior written notice with reasons, and so does any change to contract terms, which means your provider app needs a notice and acknowledgement flow rather than a silent account state change. And every aggregator must publish a mechanism through which a worker can ask how automated systems produced their fares, earnings and customer feedback, and answer within five working days. That is an explainability requirement on your dispatch and pricing services: log the inputs to every automated decision, keep them retrievable per worker, and make them readable by a human support agent.
Build this in and it is a schema decision. Retrofit it and it is an archaeology project.
What the market rewards
Public numbers show where the margin actually sits. Swiggy's quick commerce contribution margin was minus 0.2% of gross order value in the quarter ended 30 June 2026, having improved 440 basis points year on year and crossed break-even in May 2026, against a peak loss of 5.6% of GOV in the March 2025 quarter. Its food delivery segment ran at 3.1% adjusted EBITDA on GOV. Eternal reported food delivery adjusted EBITDA of ₹532 crore at 5.5% of net order value for the quarter ended 31 March 2026, and guides to 5% to 6%.
Those are single-digit margins on enormous volume. An on-demand platform earns its money in fractions of a rupee per order, which is exactly why the map bill, the take rate and the payout fee decide whether the model works. The real cost is rarely the code.
India's platform workforce was around 1 crore in 2024-25 and is projected at 2.35 crore by 2029-30, while only 3.37 lakh platform workers had registered on e-Shram out of 30.98 crore unorganised workers as of 3 August 2025. Supply is not the constraint. Retention of good providers is.
How we build
1. Discovery, 1 to 2 weeks. Supply model, city coverage, fare or fee structure, payout partner review, and a compliance map against the aggregator guidelines and the State welfare laws that apply to you. We write down what version one deliberately excludes.
2. Architecture, 2 weeks. Dispatch design, location ingestion rate, the maps call budget above, the ledger model, and the audit schema for fares and automated decisions. The ledger is designed before the screens.
3. Build, 10 to 18 weeks. Two-week increments with a working build at the end of each. Customer app, provider app, operations console and dispatch service, with the console in scope from the first increment. Framework choice is argued on maintenance cost, and our React Native versus Flutter hiring decision framework sets out how we decide.
4. Test, 3 weeks, overlapping. Load on the dispatch loop, not just the API. Battery and background location behaviour on mid-range Android. Payout reconciliation against a synthetic week of trips. Cancellation and refund paths walked end to end.
5. Pilot and stabilise, 4 to 6 weeks. One city, real providers, daily operations review, and a defect budget reserved for the first month, because live supply finds things no test finds.
A first production release across all four surfaces typically lands in the 16 to 24 week range. Anyone quoting eight weeks is quoting a clone script, and a clone script has no ledger.
Engagement models
We work in three ways, and we quote after discovery rather than before it. A fixed-scope pilot suits a single city with a defined supply type and a fixed feature list. A dedicated squad on a monthly retainer suits a platform expanding city by city, where scope changes faster than a contract can. A build and transfer engagement suits a funded operator who intends to hire an in-house team within a year and wants the codebase and the runbooks handed over cleanly. What we will not do is quote a number before seeing your supply model, because the honest gap between a 200-provider single-city service marketplace and a multi-city ride platform with its own ledger is wide enough to make any advance figure marketing. Our comparison against offshore alternatives is in India versus US app development cost.
The stack
Provider and customer apps in Kotlin with Jetpack Compose and Swift with SwiftUI where battery and location behaviour decide the product, Flutter or React Native where a shared codebase genuinely reduces total effort. Dispatch and pricing as separate services with their own datastores. Location ingestion through a streaming layer rather than a request per ping. India-region AWS, Azure or Google Cloud, with data residency decided at architecture stage. Payments through a licensed payment aggregator with escrow settlement, and UPI-first collection, which we cover in our UPI multi-PSP checkout resilience work. Store submission planned against the current Android target API 36 deadline.
Why eCorpIT
We are a senior-led engineering organisation founded in 2021, headquartered in Gurugram, assessed at CMMI Level 5 and certified to ISO 27001:2022, and an MSME registered company. We are partners with AWS, Microsoft, Google and Shopify.
You get full IP and code ownership from day one, including repositories, infrastructure definitions and runbooks. The people who scope your build write it. We design applications aligned with DPDP requirements rather than claiming a certification we do not hold.
Related work
Neighbouring pages are our marketplace app development service, ecommerce app development company for catalogue-led builds, and D2C mobile app development. Buyers who want a partner they can visit should read our mobile app development company in Gurgaon page, and anyone still shortlisting will get value from how to choose a mobile app development company.
How eCorpIT can help
eCorpIT builds on-demand and marketplace platforms for Indian and global operators: dispatch and matching, live tracking, pricing engines built to the aggregator guidelines, provider payout ledgers that reconcile against escrow settlement, operations consoles, and the audit and explainability layers the State gig worker rules now require. Our senior engineering teams work to CMMI Level 5 process discipline and ISO 27001:2022 controls. For a scoped estimate against your supply model and city plan, contact us.
FAQ
References
- Karnataka Platform Based Gig Workers (Social Security and Welfare) Rules, 2025, gazette notification
Last updated 15 August 2026.
