Nobody's first month looks like this.

Orders arrived as WhatsApp messages and got written into a ruled notebook. Payment was a UPI screenshot the buyer sent, which she cross-checked against her bank SMS. Stock was whatever she remembered pouring that week.

It worked at fifteen orders a month. At sixty it started failing in specific ways: two customers charged the wrong amount, one order shipped twice, one never shipped at all. Nothing catastrophic. Just a slow tax on every day.

The actual reason she moved was not any of that. A corporate buyer asked for a GST invoice and a link to order from. She had neither.

The store itself existed in under a minute — sign in with the WhatsApp number, name the brand, pick a theme. That part is genuinely as fast as advertised.

The ninety minutes went where they always go: photographs. Eleven products, each needing a shot that did not have her kitchen counter in the background. The AI-drafted descriptions took two minutes to write and twenty to edit into her own voice.

What she did right on day one: turned on all six WhatsApp automations before the first order, so she never built the habit of answering status questions by hand.

She posted the link once to her WhatsApp status and once on Instagram. Six orders in the first week — all from people who were already customers. Nobody new found her. That is normal and it is worth saying out loud: a store link does not generate demand, it captures demand you already have.

The problem was COD. Four of the six chose cash on delivery because it was there and it was default. One of those four refused the parcel. She had shipped ₹1,200 of candles and paid both legs of the courier for nothing.

Three settings changed, in this order. Phone verification on COD orders — the buyer confirms via OTP before the order is accepted, which removes the accidental and the fake ones outright. A ₹40 COD handling fee, visible at checkout, which quietly pushed about a third of buyers to prepaid. A ₹2,500 cap above which COD is simply not offered.

Refusals went from one in four COD orders to roughly one in fourteen over the next two months. Revenue reached ₹94,000 in month three — most of it the same customers ordering more often, because the store made reordering a two-tap operation.

Nine months in: about ₹3.1 lakh a month, roughly 60% prepaid, one part-time helper for packing, and the corporate account that started the whole thing now ordering gift sets four times a year.

The number she cares about is not the revenue. It is that she stopped answering "where is my order" entirely. That was an hour a day. Nine months of that is about 270 hours — a full working month and a half returned to her.

400 orders a month, ₹4.6 lakh in gross sales, and less cash at the end of the month than the month before. He assumed it was courier rates. It was not.

Of those 400 orders, 310 were cash on delivery — normal for groceries. Of the 310, 87 came back undelivered. Each one cost him forward shipping, return shipping, and in about a fifth of cases a product too damaged to resell.

28% RTO on a category with thin margins is not a leak. It is the business.

Sorting the 87 refusals by reason changed the plan completely:

31 — buyer unreachable on the delivery call.
24 — "I didn't order this" (impulse orders at 1am, forgotten by Tuesday).
18 — address incomplete or wrong pincode.
9 — genuinely changed their mind.
5 — courier never actually attempted.

Only nine of eighty-seven were real customer decisions. The other seventy-eight were preventable at checkout.

1 · COD phone verification. An OTP before the order is accepted. Killed most of the 31 unreachable and a good share of the 24 denials — a fake number cannot confirm.

2 · Pincode serviceability at checkout. Unserviceable pincodes are blocked before payment instead of failing at the doorstep. Most of the 18 address failures.

3 · A ₹49 COD fee, waived on prepaid. Prepaid share went from 22% to 41% in six weeks. Prepaid orders essentially do not RTO.

4 · WhatsApp shipping updates. A buyer who has read three messages about a parcel is dramatically more likely to answer the delivery call.

Orders dropped from 400 to 341 in the first full month. The COD fee and the OTP step turned some people away — and this is the part sellers panic about and reverse too early.

But RTO fell to 12%, then to 9%. 341 orders at 9% RTO delivers 310 parcels. 400 orders at 28% delivers 288. Fewer orders, more delivered revenue, materially less courier spend, and a warehouse no longer full of returned rice.

Net cash improved by roughly ₹61,000 a month on lower topline. Topline is not the score.

One catalogue could not serve all three audiences. Retail buyers must not see wholesale rates. The wedding store needs to appear in October and disappear in February without dragging the main brand's SEO with it.

Three separate stores under one login: separate catalogues, separate themes, separate payouts, separate analytics — one store picker at the top-left of the dashboard. Not three accounts, three logins and three sets of everything.

During the first Diwali, the three stores did about 900 orders in eleven days. Everything technical held. She did not.

Every refund approval, every stock question, every "can we ship this to Coimbatore" came to her, because she was the only login. Two staff were packing at full speed and stopping every ten minutes to ask her something.

She missed 40 orders' worth of stock going negative on one SKU, because nobody else could see inventory.

Five seats, scoped by role. The packing team can see orders and mark them shipped, and cannot see revenue, edit prices or issue refunds. The catalogue person can edit products in one store only. The accountant sees payouts and nothing else.

Two warehouses were split by region so the Bengaluru team stopped picking stock that physically sat in Surat. Inventory became a fact everyone could read instead of an answer only she could give.

The following season did roughly 1,400 orders across the three stores with the same five people. Nothing about the platform changed. What changed was that four other people could answer the questions that used to queue behind her.

Her measure of the year: she approved 11 refunds instead of 300.