How to sell food online in India: licences, stack and the first 90 days

Before anything: the legal minimum, in order

Week zero paperwork, in dependency order: FSSAI registration (Basic tier to start; upgrade to State Licence as turnover or platform requirements demand — the full tier logic is in the compliance checklist); GST registration (mandatory from the first rupee if marketplaces or quick commerce are anywhere in your plan); a current account and basic accounting setup; trademark application filed early (marketplace brand registries accept applied-for marks, and the queue is long); and labels designed to regulation before the first print run — reprinting packaging because the nutrition panel was wrong is the most avoidable ₹50,000 in the industry. None of this needs consultants at launch scale; it needs a competent CA and one careful week.

The minimum viable stack (resist the tool buffet)

What you actually need to take orders: a Shopify store (or equivalent — the comparison is in the platform piece) with a fast, mobile-first theme; a payment gateway with UPI intent flow (the gateway comparison); a courier aggregator account for day-one pan-India shipping (the courier piece); WhatsApp Business API via an entry-tier provider for order updates; and the Meta pixel + Conversions API wired before the first ad rupee. That is five subscriptions, not fifteen. Every additional tool before product-market fit is a distraction with a monthly fee.

Days 1–30: sell ugly, learn fast

The first month's only job is honest signal. Launch to the warmest circle you have — family, friends, colleagues, the founder's own network and neighbourhood groups — at full price (discounted feedback is polluted feedback). Ship every order with a personal note and a WhatsApp follow-up asking two questions: what made you try it, and would you reorder? Fix what the first twenty orders teach — the pack that leaks, the courier that delays, the description that oversells. Collect every buyer into the WhatsApp list with consent. Do not run paid ads yet; at this stage ads amplify ignorance. Target: 50–100 orders, a complaint list, and five genuine review paragraphs with photos.

Your first hundred customers are research wearing a revenue costume. Price them that way.

Days 31–60: the first paid experiments

Now buy signal deliberately. ₹1,500–3,000 a day on Meta, broad targeting, six to eight genuinely different angles built from the exact words your first buyers used (the method is the positioning-test playbook at starter scale). Judge angles on CAC and the post-purchase "why did you buy" answers, not clicks. In parallel: the PDP gets rebuilt weekly as objections surface — delivery time, taste risk, price-per-use maths; the replacement-first returns flow goes live (the policy piece); and the review ask fires on WhatsApp at day three post-delivery. Target by day 60: one angle with a repeatable CAC you can afford at your contribution margin, and 25+ photo reviews banked.

Days 61–90: double down or diagnose

Two honest paths. If an angle works: scale it 20–30 percent a week, launch the trial-pack or bundle that fixes your first-order economics, switch on abandoned-checkout and replenishment flows, and prepare the marketplace listing with your proven creative and banked reviews — the full rollout logic is in launch D2C-first. If nothing repeats: the data says product, price or positioning — not "more budget". Re-read the first-buyer interviews, test a different pack size or price point, and consider whether the category's real demand lives in a different angle entirely. Ninety days of disciplined direct selling answers questions two years of marketplace rank-chasing never will.

The five mistakes that kill food brands in year one

1. Marketplace-first launch — velocity without learning, reviews on rented land, price wars before brand exists.
2. Discount-led acquisition from day one — trains the market that the MRP is fiction before the MRP has meant anything (the treadmill starts here).
3. Ignoring unit economics until "later" — contribution margin per order, computed before the first ad, is the number every other decision hangs on.
4. Shipping ops as an afterthought — one festive week of broken deliveries can bury a young brand's ratings permanently.
5. Fifteen tools, no list — the only asset that compounds from day one is the owned audience; most launch stacks optimise everything except it.