पाठशाला Pathshala · वृद्धि Vṛddhi, Growth · Lesson 03 · Start

Finding the first channel that works

Most companies grow through one channel, and nobody knows in advance which. Run six cheap tests in a month, measure the cost of a qualified user from each, and put the next quarter behind the winner.

Pathshala, The Founder Library · 11 October 2026 · 7 min read

A founder in Jaipur runs Instagram ads because the last company she worked at ran Instagram ads. Her product sells to small hotel owners, who are on WhatsApp all day and on Instagram never. Six months of spend later she has a brand page and no customers. She did not pick the wrong channel. She did not pick at all.

Almost every company that reaches scale does so through one dominant channel, and the founders of those companies will tell you they did not know which it would be until they had tried several. This lesson is the method for trying several in a month without spending the runway on it. It borrows its structure from Gabriel Weinberg and Justin Mares’ Traction and adds the Indian channel list, the Indian costs and a scorecard you can fill in from your own numbers.

Why one channel, and why you cannot guess it

Weinberg and Mares catalogue nineteen traction channels, from viral loops and search engine marketing to trade shows, speaking and community building, and their central observation is that at any stage one of them does most of the work. The job is to find out which. Their method, which they call the Bullseye, is three rings: brainstorm how the company could use every channel, pick the three most promising, run cheap tests, and then focus on the one that performs. In a 2015 interview with Mixergy Weinberg put the budget for a test at not longer than a month or about a thousand dollars, and the mistake he warns against is the one the founder in Jaipur made: defaulting to the channel most common in your industry, when the channel that works for a new company is usually one its competitors have left under-used.

The deeper reason you cannot guess is that the channel is not a marketing decision. Brian Balfour’s essay on the four fits argues that products are built to fit channels and not the other way round: a product that spreads through WhatsApp forwards has to produce something worth forwarding, a product sold through LinkedIn has to be explainable in three lines to a stranger, a product bought from a Google search has to answer the question the buyer typed. The channel shapes the product. That is why the test has to be run early, while the product can still be shaped.

The six channels to test first in India

The nineteen are the full menu. For a company at its first hundred users in India six of them are worth a month each, and they are chosen because together they cover the three ways a buyer here first hears of anything: from a peer, from a search, or from someone who turned up. WhatsApp communities and groups: the trade group, the society group, the alumni group, the group the industry already runs. Not a broadcast list; a founder present in the group, answering, and offering the product when the problem comes up. LinkedIn outreach: for anything sold to a company. India is one of LinkedIn’s largest markets and the decision-maker in a two-hundred-person firm in Pune is reachable by name. Meta ads: Instagram and Facebook, the default channel for consumer and D2C, and the one whose costs rise fastest as you scale. Google search ads: for products people already know they need and type in, from GST software to a CA in Noida. Events and trade bodies: the association meeting, the industry expo, the Chamber’s monthly lunch, the stall at the mandi. Referrals and partners: an existing customer introducing the next, a CA or a distributor or a software reseller who already has the relationship.

Leave out, for now, anything that takes more than a month to show a signal: SEO, content, PR, app-store optimisation. They may be the eventual channel, and they are in the book, but they cannot be tested in thirty days and this lesson is about thirty days.

How to run a test that answers three questions

Weinberg says every test should answer three questions: what does a customer cost from this channel, how many customers are available through it, and are they the right customers. Hold the method to those three and the test designs itself. A fixed budget per channel, say ₹10,000 to ₹50,000 depending on the channel and the stage, and a fixed month. A written definition of qualified before the test starts: a lead is qualified if it fits the one-sentence segment from your [customer interviews](/library/the-customer-interview-done-properly) and did the first real thing, whether that is a demo booked, a first order placed or a first week of use. Sign-ups are not qualified. Downloads are not qualified. Founder time counted as cost, because a WhatsApp community that takes forty hours of a founder’s month is not free and a Meta campaign that takes four hours is not expensive merely because it has an invoice. One channel, one owner, one sheet. Every lead tagged with where it came from, by hand if necessary.

The number at the end is the all-in cost of one qualified user: spend plus founder hours at an honest rate, divided by qualified users. It is the earliest form of the paid CAC that a16z’s 16 Startup Metrics says investors want to see broken out by channel, and the same note gives the warning that goes with it: costs rise as you reach further into a channel, with their example being a dollar per user for the first thousand and five to ten dollars for the next hundred thousand. A channel that is cheap at ₹200 a user in its first month may not be cheap at scale. That is the second question, how many are available, and the test answers it only roughly. Watch whether the cost rose inside the month.

Reading the scorecard

The cheapest bar wins, but read the whole chart before deciding. Fit to the price. Balfour’s channel-model fit is the constraint founders forget: the cost of a channel has to sit well below what a customer pays. A product at ₹499 a month cannot be sold through events that cost ₹1,500 a qualified user and a six-month payback; a product at ₹5 lakh a year can afford a founder’s month on LinkedIn. Run the winning channel’s cost through the [CAC, LTV and payback](/library/cac-ltv-and-payback-the-three-numbers) arithmetic before committing to it. Ceiling. A referral channel may be the cheapest and also run out at the edge of your current customers’ address books. The channel to double down on is the cheapest one that could plausibly produce ten times this month’s qualified users at under twice the cost. Quality after the month. Go back to the qualified users from each channel thirty days later and see who is still there. A channel that produces cheap users who leave in week three is the most expensive channel on the chart.

A channel is not a place to put money. It is a place where a particular buyer is already paying attention, and the test is whether you can be worth it.

Doubling down, and what it does to the product

Once a channel has won, the next quarter is the winner’s. Weinberg’s rule of thumb is to spend half the company’s effort on product and half on traction, and most founding teams spend ninety per cent on product; the correction is to give the winning channel a named owner, a weekly number and the majority of the traction half. Put roughly eighty per cent of the growth effort there and keep one secondary channel alive at twenty per cent, because every channel saturates and the second one is where you will be standing when the first does.

Then let the channel shape the product, which is what Balfour means by fit. If WhatsApp communities won, the product needs a shareable artefact: a receipt, a report, a scorecard, something a user forwards without being asked. If LinkedIn won, the product needs a demo that runs in fifteen minutes and a one-page case study with a named customer. If Google search won, the landing page becomes the product’s most important screen and the words on it are the words buyers typed. If referrals won, the referral has to be built into the product rather than asked for in a message. The growth rate the company reports every week, the number that [matters more than any other early](/library/growth-rate-is-the-only-number-that-matters-early), will from here on be mostly the growth rate of this one channel.

A monthly ritual

The first month runs all six tests at once, which is heavy and is also the only way to compare them under the same conditions. After that, keep the scorecard as a standing document and update it on the last Friday of every month with three rows: the winning channel’s cost per qualified user this month against last, the secondary channel’s, and one new test. One new test a month, at the same capped budget, is how the next channel is found before the current one is exhausted. When the winner’s cost has risen for three consecutive months, the next test moves to the front of the queue. When a test beats the winner for two months running, swap them. Write down the date and the number each time, because the scorecard is also the record that will one day show an investor that the company knows where its customers come from and what they cost, which is most of what they will ask.

The method is Weinberg and Mares’; the channel list and the costs are India’s and will change. The sources are below. Start the six tests on the first of the month.