Solutions · Food & grocery
Grocery delivery app: what to decide, build and test.
Planning one? Here's what to settle before you brief anyone, what goes in the first version, how the build runs week by week, how the beta is tested with real people, and the numbers that show it's working.
Dark store, kirana network or your own shop: pick the stock model
A grocery app lives or dies on stock accuracy, not screens. If the app says an item is available and the shelf is empty, every order becomes a substitution or a refund. So the first decision is where the stock sits.
- Dark store. A closed warehouse you stock and staff. Picking is fast and stock is yours to count, but you pay rent, staff and wastage before the first order, and it needs enough orders in a small radius to pay off.
- Your existing supermarket or kirana. Cheapest start. Stock is shared with walk-in customers, so the app needs a live sync with your billing, or pickers will find empty shelves.
- A network of neighbourhood shops. Needs less capital and suits smaller cities. Zippit lets one cart span shops within a radius, with a base fee of ₹25 and ₹15 per extra shop, at most four shops, and a waiver above a basket value. The catch is that you do not control each shop's stock or packing time.
- Ten-minute versus slotted delivery. Ten minutes needs dark-store density. Slots such as a morning or evening window work from a normal shop and cost much less to run.
The hyperlocal delivery guide and the grocery delivery playbook go through the economics. Decide these before briefing:
- Stock model, and who updates stock counts and how often
- Delivery radius and promised time, instant or slot
- Catalogue size at launch, and where product data and photos come from
- Own riders or partner, and who packs the bag
- Substitution policy: auto-substitute, ask the customer, or refund
- Minimum basket and delivery fee rules
Version 1 for a grocery app: catalogue, picking and substitutions
| Module | First version | Add later | Why |
|---|---|---|---|
| Catalogue | Top 1,000 to 2,000 items with photos, units and MRP | Full range, brand pages | A small clean catalogue beats a large wrong one. |
| Search | Hindi, Hinglish and English spellings, brand and unit | Voice search, typo learning | People type doodh and pyaaz, not what your database calls them. |
| Stock sync | Counts updated from billing or a daily upload | Real-time inventory with low-stock alerts | Wrong stock is the main cause of refunds. |
| Picker screen | Pick list sorted by aisle, tap to confirm, not-found button | Barcode scanning, pick-time analytics | Sorting by aisle is where picking speed comes from. |
| Substitutions | Suggest a match, customer approves in the app | Preference memory per customer | Silent substitution is the fastest way to lose a customer. |
| Serviceability | Pin code check before the cart | Dynamic radius by rider supply | Never let someone fill a cart you cannot deliver. |
| Fees and bill | Delivery, small-cart and handling fees, each explained | Surge and rain fees | Fees without reasons look like tricks. |
| Payments | UPI, cash on delivery, refund to source | Wallet, autopay | Refunds are frequent in grocery. Make them easy. |
| Rider app | Bag pickup, drop, proof of delivery | Multi-stop routes across shops | Multi-stop only matters for a shop network. |
The ten to twelve week plan, with the data work you own
| Weeks | We build | You supply |
|---|---|---|
| 1–2 | Order flow, fee rules, aisle structure, clickable prototype | Store layout, product categories, delivery area and time promise |
| 3–5 | Shopper app, catalogue, search, cart | The product master sheet: names, units, MRP, photos, local spellings |
| 6–8 | Picker screen, rider app, substitutions | Two or three pickers and riders for testing, Android phones for them, packing materials |
| 9–10 | Payments, ops console, stock sync with your billing | Gateway account and KYC, access to the billing or inventory export |
| 11–12 | Dry-run orders in the real store, fixes, store listings | FSSAI details, Play Store account, staff shifts for launch week |
How we work with you each week
The same rhythm runs from week 1 to launch, so you always know what's happening and what we need from you.
- Friday: a demo link and a 30-minute call. You open the latest build on your own phone and try it yourself, rather than watching our screen.
- Within 48 hours: your feedback, in writing. In one shared doc, next to the screen it's about. Written feedback is clear, dated and doesn't get lost between calls.
- Monday: triage. Every comment goes into fix now, next week or later. Anything that changes the agreed scope gets a written quote before we touch it.
- Each milestone: two revision rounds, then your sign-off. Sign-off is against the scope we agreed, so it's clear what done means.
- One decision-maker on your side. Most delays aren't build time. They're feedback waiting on several people who disagree.
The full version, including how the beta and user testing work across every kind of project, is on our process page.
The beta: one store, one pin-code cluster
The beta starts around week 10, once stock sync with your billing works and a real order can go from cart to picked bag to doorstep. It runs from the one store you will launch with, through the dry-run weeks of the plan.
- What's in it. The launch catalogue, search, cart with the fee breakdown, UPI and cash on delivery, the picker screen with the not-found button, substitution approval, the rider app and the ops console.
- What stays out. Autopay and weekly orders, a wallet, barcode scanning, rain fees and a second store. Each adds noise before you know the basic loop holds.
- Who uses it. Two or three of your pickers, two riders on their own Android phones, and 10–15 households in the first pin codes: people who already buy from your shop or order on WhatsApp. You recruit them. Relatives who never shop with you will not show you a real weekly basket.
- How they get it. Households install through TestFlight or Google Play testing. Pickers and riders get the Android build on the phones they will use in the store. Ops uses a staging link to the console.
- How long. Two weeks, so each household places at least two real orders and a weekend rush is included. Every screen has a feedback button that sends a screenshot and a note to a shared sheet.
A new personal Google Play developer account has to run a closed test with at least 12 testers for 14 days before it can publish. Count the beta group towards that, or use an organisation account.
What we test with real people, and the tasks we set
Each session is one task, done on the tester's own phone and in the words they normally shop in. We sit beside them and say nothing until they finish or give up. Around five people per role is enough to show the main problems. We fix those, then run the same tasks again.
| Who | The task | What we watch for |
|---|---|---|
| Household shopper | "Buy 2 litres of toned milk, 1 kg pyaaz, a 5 kg atta bag and a packet of instant noodles for tonight, and pay by UPI." | Does search find pyaaz and atta typed in Hinglish? Do they pick the right pack size? Is the small-cart fee clear before they pay? |
| Household shopper | "Your usual atta brand is out. The app suggests another. Accept it or ask for a refund." | Do they notice the substitution request in time? Is the price difference shown? Does the refund go back to UPI without a call? |
| Picker | "Pick this 18-item order in the storeroom, and mark the paneer as not found." | Does the aisle order match the real shelves? Does the screen keep working where the signal drops? How long does the not-found step take? |
| Rider | "Collect two bags at the counter and deliver to the third floor of a building with no lift, cash on delivery." | Do they get the right bags? Can they mark cash collected and give change without the app getting in the way? |
| Store owner or ops | "A customer says the eggs arrived broken. Refund the eggs only, and check today's stock mismatches." | Can they refund one line, not the whole order? Do the stock gaps from pickers show up where they look? |
The numbers that say it's working
In week 1 we agree on two or three of these with you. If you take phone or WhatsApp orders today, we note the current figure as a starting point. The tracking goes in before the beta. We report the numbers at launch and again a month later. We can't promise where they land, but you will see them clearly and can act on them.
| Goal | What we measure | Where it comes from |
|---|---|---|
| The app shows what's on the shelf | Items marked not found per 100 picked | Picker screen |
| Substitutions don't cost customers | Share of substitution requests accepted, refunded or left unanswered | Order data |
| Search finds local words | Searches that return nothing, with the words people typed | App analytics |
| Bags leave the store quickly | Time from order placed to bag handed to rider | Picker and rider app timestamps |
| Households keep ordering | Households that place a second order within 7 days | Order data |
| Fewer bills need fixing | Refunds and complaints per 100 orders, by reason | Ops console |
Grocery app cost: where it lands and what shifts it
On the app cost guide, a shopper app, picker screen, rider app and ops console is a two-sided product with an admin side, ballpark ₹10–25 lakh from a small studio. A multi-store network with live maps and payments sits in the ₹25–60 lakh band. These are ballparks; the studio quotes on your brief.
- Billing and inventory integration. Connecting to your existing billing software is often the biggest unknown. A daily upload is cheap; live sync is not.
- Catalogue size and change rate. Prices and stock that change hourly need more engineering than a weekly price list.
- One store or several. Each extra store multiplies stock, picking and settlement logic.
- Live tracking and maps. Status updates are light. A moving map costs more to build and to run.
Packaged food, MRP, refunds: the rules a grocery app trips on
- FSSAI. Food sellers and platforms need the right FSSAI licence or registration, and the number belongs on invoices and receipts. FSSAI has directed platforms not to list sellers without a valid one. Which category applies depends on size and model, so check it on the FSSAI portal before launch.
- GST by item. Rates differ by product, and many fresh items are treated differently from packaged ones. Rates and the tax treatment of platform delivery fees can change. Store the tax rate per item, not one rate for the cart, and have a CA confirm the rates for your catalogue.
- MRP and packaged goods. Packaged commodities carry mandatory declarations such as MRP. Show them on the product page. Confirm the current Legal Metrology requirements for online sellers with a lawyer.
- Complaints and refunds. The Consumer Protection (E-Commerce) Rules expect a named grievance officer who acknowledges within 48 hours and resolves within a month. The rules were amended in 2026, so read the current text.
- Collecting payments. If you collect money for several shops and pass it on, RBI's Payment Aggregator Directions of 2025 apply to the entity doing that. Use a licensed gateway.
- Recurring payments. For autopay or weekly orders, RBI's e-mandate framework lets recurring debits up to ₹15,000 per transaction go without extra authentication, and requires a notice at least 24 hours before each debit.
- Data. Addresses and order history are personal data. The DPDP Rules phase in over 18 months from November 2025. See the DPDP checklist.
This is general information for planning, not legal or tax advice. Rules and rates change; confirm with a CA or lawyer.
Seven grocery mistakes, and the fix for each
- Promising ten minutes without density. A promise you miss twice is worse than a slot you hit. Offer slots until orders per area justify instant.
- Importing a messy catalogue. Duplicates and missing units cause wrong bags. Clean the master sheet before the build.
- Showing stock you do not have. Sync from billing, or show less.
- Auto-substituting. Customers cancel when brand or size changes without asking. Ask, in one tap.
- One tax rate for the whole cart. The bill can be wrong on every line. Tax per item.
- Ignoring weak networks. Pickers work in storerooms with poor signal. Make the picker screen work offline and sync later.
- Free delivery on every order. Small baskets lose money. Set a minimum and say why.
A safer start is one store, one pin-code cluster and a list of 1,000 items, run for a few weeks before you widen the radius or add stores. See also regional language apps for search in local phrasing.
Cost and timeline
10–12 weeks for a first version.
Fixed quote within 48 hours of your brief: scope, timeline and price in writing before we start. These are what move the number.
- Single store, many stores or dark stores
- Catalogue size and how often prices and stock change
- Your own riders versus a delivery partner
- Integration with your billing or inventory system
- Number of languages for shoppers and staff
Questions
About your grocery delivery app.
What does a grocery delivery app cost in India?
The cost depends on the number of stores, the size of the catalogue and how many apps you need: shopper, picker, rider and ops. We send a fixed quote within 48 hours of your brief. The app cost guide explains each driver so you can plan before we talk.
Can it connect neighbourhood kirana shops instead of a dark store?
Yes. Our concept build Zippit does exactly that: shops within a short radius share one cart, and a single runner collects from each on a multi-stop route. It needs less capital than a dark store and suits Tier 2 and Tier 3 cities.
How long until we can launch?
A full version with shopper, picker, rider and ops screens usually takes 10 to 12 weeks. A store with its own delivery staff and a simpler catalogue can launch sooner. A two-week concept sprint gives you a clickable prototype first.
Will it work on cheap phones and patchy networks?
Yes. We keep the catalogue cached, compress images and test on low-end Android handsets. The rider and picker apps keep working through weak signal and sync once the connection returns. Prices and stock refresh in the background, so a shopper never waits on a spinner to see what is available.
Do we own the code?
Yes. You get the full source code, design files and brand system at hand-over. Your own engineers or any other team can carry on from there with no licence fees. We also leave documentation on how the order flow works.
Read next
- DeliveryStarting a hyperlocal delivery business: the apps you need →The apps behind a hyperlocal delivery business in India: customer app, store or picker screens, rider app and ops console, plus models for groceries, dairy and parcels.
- CostHow much does it cost to build an app in India? →Ballpark app development costs in India for 2026, from a simple app to a multi-sided marketplace — what moves the price, and how to get a quote you can trust.
- LanguagesBuilding apps in Hindi, Tamil, Kannada and other languages →How to build an app or website in Indian languages: scripts and fonts, longer text, voice input, transliteration, lakh and crore formats, and translation that sounds right.
Start a project
Tell us how your business runs. We'll build what it runs on.
Four taps, about a minute. Fixed quote within 48 hours.


