पाठशाला Pathshala · वृद्धि Vṛddhi, Growth · Lesson 02 · Start

Do things that do not scale: the India edition

Paul Graham’s advice was to recruit the first users by hand. In India the hands are a shop visit, a trade market before it opens and a one-to-one WhatsApp message. The method, and what to write down.

Pathshala, The Founder Library · 11 October 2026 · 9 min read

The app is live. The Instagram page has a logo. Three weeks later there are forty sign-ups and eleven of them are friends. Nothing is wrong with the product. The founders are waiting for users to arrive, and users do not arrive. They are fetched.

In July 2013 Paul Graham wrote Do Things that Don’t Scale and it has been the most-assigned essay at Y Combinator since. Its argument is that almost every startup needs a push to get going and that the push is made of work founders would be embarrassed to describe to an investor: recruiting users one at a time, installing the product for them, writing thank-you notes by hand. This lesson takes the essay as read and does the translation. What the manual push looks like in Bengaluru or Indore rather than Palo Alto, how WhatsApp changes it, and what to write down while doing it so that the hundred users you fetch by hand become the method that fetches the next ten thousand.

What Graham actually said

Four moves, all of which still hold. Recruit manually. Nearly all startups have to go out and get their first users; the ones that seem to have launched into instant demand almost always had a founder doing this quietly. Stripe’s founders, when someone at a YC dinner agreed to try the product, did not send a link. They said “right then, give me your laptop” and set it up on the spot, a move YC still calls the Collison installation. Go where the users are. Airbnb’s founders flew to New York, went door to door to the hosts, took better photographs of their flats and learned what hosts needed. Graham’s view is that about thirty days of that work was the difference between the company existing and not. Delight. Early users should be made to feel that signing up was one of the best decisions they ever took; Wufoo sent hand-written thank-you notes to every new user for as long as it could. Do it by hand first. Stripe’s early “instant” merchant accounts were set up manually behind the scenes. Viaweb built online stores for merchants who would not use the software themselves and in doing so learned what the software needed to be.

The reason founders resist all of this is that it feels small. Graham’s answer is arithmetic. A hundred users is nothing. A hundred users growing ten per cent a week is about fourteen thousand a year later. The manual work is not meant to be the growth. It is meant to build the base that grows. Treat the figure below as the point of the lesson: move the hand-recruited number and watch what happens to the total only once the base starts bringing its own.

The India edition, in person

India’s first advantage is that the users are physically findable. A product for kirana owners has its users standing behind counters from eight in the morning. A product for traders has them at the wholesale market before it opens and at the trade association meeting on the first Tuesday of the month. A product for parents has them at the school gate at two in the afternoon and at the society notice board on Sunday. A product for small manufacturers has them in a named industrial estate with a gate register. Nowhere in the world is it cheaper to stand next to a hundred potential users in a day, and most founders spend that day on a landing page instead.

Zomato is the Indian founding story that fits Graham’s essay exactly. Deepinder Goyal and Pankaj Chaddah were consultants at Bain in Delhi whose office café kept a set of stapled restaurant menus so nobody could walk off with them. In 2008 they scanned about fifty of them and put them on an intranet page; traffic inside the firm convinced them it was a business, and the early site was literally scanned menu cards, uploaded by the founders, as Goyal told Knowledge at Wharton in 2012. Four years in, more than half the company still worked on gathering content by hand. It did not scale. It was the product.

The rules for in-person recruiting are the rules of the customer interview with a sign-up at the end. Go as a founder, never a hired promoter; the judgement about what to build is formed at the counter. Carry the product on your own phone and set the user up there and then: install it, create the account, enter their first real data, show them the one thing it does for them. Leave your personal number, not a support line. Write down the name, the shop, the time it took and what they said, before you move to the next shop. The Collison installation translates without loss; the only Indian addition is that the first question is usually “who else is using it”, and the honest answer in the first week is “you are the third, and here are the other two”.

The India edition, on WhatsApp

WhatsApp is where the manual push happens when the user is not in front of you, and it is where most founders get it wrong in the same way: they treat it as a broadcast channel when its value at this stage is that it is a conversation. A one-to-one message from a founder, by name, to a person who has given their number, referring to something specific about their shop or their job, gets a reply. A forwarded flyer in a group of two hundred gets the group muted.

The rules of the platform are the rules of good manners made enforceable. Meta’s WhatsApp Business Messaging Policy says a business may contact people only if they have given their number and opted in to receive messages, that users should expect the messages they receive, and that people can block or report a business, after which “our systems will limit the amount of messages a business can send”. On the API, a business may open a conversation only with an approved template and may reply freely only inside a twenty-four-hour window after the user’s last message. None of this constrains a founder doing the work properly. The user you met at the counter gave you their number and expects to hear from you; the person whose number you scraped from a directory did not, and the block button is how they say so.

Run it as a daily discipline rather than a campaign. Every new user gets a personal message within an hour of signing up: thank you, here is my number, tell me the moment anything breaks. Every user who has gone quiet for three days gets one question, not a nudge: did it do what you needed on the day you tried it? Every reply is answered by a founder inside the hour, during working hours, and the answer is written in the user’s language, which in most of India is not English. The WhatsApp Business app, as against the API, exists for exactly this stage: a business profile, quick replies, labels, a catalogue, and a founder typing. Keep the group for later. For the first hundred users the group is a place where one unhappy user teaches the other ninety-nine to be unhappy before you have had a chance to fix it.

The first hundred users are not acquired. They are fetched, one at a time, by someone whose name they know.

The concierge version: do by hand what you will automate

Graham’s “manual” and “consult” moves are the part of the essay founders in India most need and most resist, because building software feels like progress and doing the job by hand does not. The logic is the other way round. Nothing teaches you what the software must do faster than doing the job yourself for ten paying customers. A bookkeeping product for small traders begins as a founder with a spreadsheet doing the books for ten traders and discovering that the hard part is not the ledger, it is getting the invoices out of the WhatsApp photos they send at midnight. A logistics product begins with the founder on a two-wheeler delivering the first fifty orders and learning which societies will not let a stranger past the gate. A B2B SaaS product begins as a weekly call in which the founder does the analysis and emails the result; the product is whatever the founder found themselves doing every Tuesday.

Two rules keep the concierge phase honest. Charge for it, at a price you would be happy to automate towards, because a user who pays ₹999 a month for a founder’s manual service is a customer and a user who gets it free is a focus group. And set a number at which you stop: when the same step has been done by hand fifty times and it has stopped changing, build it. Not before. Automating a step you have done three times means automating your guess about the step.

What to write down while you do it

The reason this is a growth lesson and not merely a hustle lesson is that the manual phase, done properly, produces the document every later phase runs from. Keep a single sheet, one row per user, from the first day. Where you found them: the market, the society, the introduction, the LinkedIn message, the group. What you said, in the form that worked, and the objection you heard, in their words. How long it took from first contact to a working account, in minutes of your time. What it cost in rupees, including the auto, the chai and the ad if there was one. Whether they are still using it two weeks later, and if not, why.

After a hundred rows that sheet answers the three questions the next lesson, [finding the first channel that works](/library/finding-first-channel-that-works), is built on: what a qualified user costs from each place you found them, how many more there are in that place, and whether they are the right users. It also becomes the script for the first salesperson and the onboarding flow for the product. Founders who skip the sheet and keep it in their heads find at month six that they cannot explain to a new hire why some users stuck and others did not, and have to go and do a hundred by hand again to find out.

The daily and weekly ritual

Close’s guide for founders, Zero to Thousands, gives the first-day rule in one sentence: in the next twenty-four hours ask at least ten qualified prospects for a conversation or a clear next step, then write down what you learned and what you will change. Keep that as the daily unit. Ten real asks a day, by a founder, in person or one to one. Not ten posts.

Weekly, on Saturday morning, count three things. Users signed up by hand this week. Users who came without you doing anything, which is the base starting to work. Hours spent per user signed up, which should fall as the script improves and will tell you when the step is ready to be built. Graham’s September 2012 essay Startup = Growth puts the target: a good growth rate during YC is five to seven per cent a week, and the rate, not the count, is what to watch. When the weekly count of users who came on their own exceeds the count you fetched, for four weeks running, the manual phase has done its job. Until then, go back to the market.

Graham’s essay takes twelve minutes to read and is listed below. Read it before Monday, then spend Monday somewhere your users are.