Two and a half years as a founder, and knowing when to stop

01

The problem

Some juniors from my college had an idea: denim retailers all buy the same white-label jeans from wholesalers in Gandhi Nagar, Delhi. If they pooled their orders, they would get bulk discounts, and the platform could take a commission. They needed a technical co-founder. I wanted to start a company, so I jumped in.

We dropped the idea within 14 days. When we spoke to more retailers, we found the trade runs on one to two months of credit, and retailers would not pay up front. An angel investor told us the same thing: we did not have the capital to absorb even one bad order.

That set the pattern for the next two and a half years: find a model that works with almost no money, and stop fast when one does not.

02

Who the users were and what I learned

Each bet taught me something about a different user:

• Retailers run on credit, so a platform that wants cash up front does not fit.
• Shoppers want it cheap or quick. Nobody would wait three to six hours to save at most about 5% against BigBasket. We can't do it cheaply, but we can do it quick.
• Society shopkeepers in Noida took orders by landline and wasted calls on items they did not have.
• Kirana sellers in Panipat had no delivery boys and would not pay a subscription for an unproven product.

03

Options considered

The company was a series of bets. Here is each one and why it ended.

1. B2B group buying for denim retailers. Stopped within 14 days: the trade runs on credit.
2. An investor's cosmetics app. About a month, until the investor said it would not work.
3. Group buying for shoppers. Stopped: the discount was too small to be worth the wait.
4. Quick delivery from a Noida society's grocery shop (kept). We rode along with the shop's own delivery boy and got residents to install the app at their door. About half the grocer's orders came through it.
5. Fight Blinkit, or move (chose to move). Blinkit arrived about a month later, with much better prices and range, and orders dropped sharply. Rather than burn money in a metro price war, we moved to Panipat, a tier-2 town with no funded competitor, within a month.
6. Subscription or commission (chose commission). We charged 5-15% per order by category, about 5% on groceries. 100+ sellers signed up.
7. Restaurants. Planned for a few months, the rebuild took seven, and restaurants brought only about 5% of orders.

04

What shipped

• All 5 apps, 3 mobile and 2 web, built alone for 12+ months: design, user stories, build, payments (Razorpay, later Paytm) and handover documentation.
• A multi-seller app with Google Maps, rebuilt from the single-shop version and launched in Panipat.
• Fill-rate fixes. Kirana stock is shared with walk-in customers and nobody tracks it, so orders failed. If a shop had 10 of an item, the app listed 5, and shops that filled more orders were listed first.
• Item-unavailable alerts. When a seller dropped an item, the buyer got a WhatsApp message, an SMS and a push alert within a minute or two, with a short window to cancel.

The pitch that won funding was not a delivery fleet. ONDC was coming, so we pitched software to bring tier-2 sellers online. That got us into AIC-ISB's SaaS cohort, then the Startup India Seed Fund. Somebody took a bet on us.

05

Result

• ₹25L raised, a team of 8, 100+ sellers, 5,500+ sign-ups and 4,800+ orders.
• Order cancellations fell 20% after the fill-rate fixes.
• Meta ad click-through rose 10-15% after A/B tests showed discount-led creatives beat assortment-led ones.
• A clean stop. In late 2023 I asked investors to back owned dark stores, to fix fill rate, and electric vehicles, to cut delivery cost. They called both long shots. I wound the company down with about a month of runway: every seller paid, the ~8 staff given a month's severance and help finding jobs, and assets returned to AIC-ISB.

The lesson I took: in a tier-2 town with no funded competitor, the limit was not price or speed. It was fill rate, at shops whose stock I did not control.

06

What I'd do differently

I would have attacked fill rate before the restaurant rebuild. Stock accuracy, not demand, was what held the business back, and seven months on restaurants did not touch it.