पाठशाला Pathshala · वृद्धि Vṛddhi, Growth · Lesson 10 · Start
The landing page and the first conversion test
Before building, put the offer on one page, send it the right people and ask for something that costs them a little. The result is the first honest number the company will have.
Pathshala, The Founder Library · 11 October 2026 · 7 min read
Forty friends said they would use it. Twelve people on a call said they would pay. A landing page sent to eight hundred people who fit the description produced three sign-ups and no payments. All three numbers are true. Only the last one was measured.
The first landing page is not marketing. It is an experiment with a hypothesis, a population, an action and a pass mark, and its job is to produce the first number in the company that was not given out of politeness. This lesson is how to write the page, how to choose an action that measures intent rather than curiosity, how to set the pass mark before launch, and how much traffic the test needs before the number means anything. It builds on the [minimum viable product](/library/mvp-that-is-actually-minimum) lesson: the page is often the smallest MVP there is.
A page that states the offer
A test page has five parts and nothing else. A headline that names the outcome for a named buyer: “GST filing for Surat textile traders, done in an hour a month,” not “The future of compliance.” Two or three lines of how: what the buyer does, what they get, how long it takes. The price, or a price range, because a page without a price measures interest in something free. One piece of proof, which before launch can be the founder’s relevant experience, a number from the customer interviews or a screenshot of the working prototype. One action, repeated at the top and the bottom. No navigation, no blog, no “about us”. Each of those is a door out of the experiment.
Write it plainly. Unbounce’s 2024 Conversion Benchmark Report, drawn from more than 41,000 landing pages and 57 million conversions, found that pages written at a fifth-to-seventh-grade reading level converted at 11.1 per cent against 5.3 per cent for professional-level writing, and that word count, reading time and difficult words all correlated negatively with conversion. For an Indian page this argues for short sentences, the buyer’s own vocabulary from the interviews and, where the buyer thinks in Hindi or Tamil, a version in that language. It also argues against the founder’s pitch deck pasted into a web page.
Choose an action that costs the buyer something
The weakest signal of intent is an email address. It costs nothing, and people give it to be polite, to be notified or to be left alone. Joel Gascoigne’s account of how Buffer was validated is the classic correction. His first page explained the product and asked for an email. When people signed up he added a page between the two: a pricing page with plans, so that a visitor had to click a plan before giving an email. The extra step tested two things at once, which plan people chose and whether they were keen enough to take one more click. Enough did, some of them on paid plans, for him to start building; Buffer launched about seven weeks later and had its first paying customer within four days.
Rank the possible actions by what they cost the buyer and pick the strongest your situation allows. An email is the weakest. A click on a priced plan is stronger. A WhatsApp message that starts a conversation is stronger again in India, because it opens a dialogue the founder can follow up. A booked fifteen-minute call with a real calendar slot is stronger still. A refundable deposit through a UPI payment link, even ₹99 or ₹499, is the strongest signal short of a sale: a person who pays to reserve has the problem, has the money and believes you. Whatever the action, tell the truth on the page about where the product is: “launching in March; reserve your place” is honest, and a fake “buy now” that ends in an apology wastes the buyer’s trust and the test’s data.
Set the pass mark before launch
An experiment without a pass mark is a story waiting to be told. Write three things down before the page goes live: the segment it will be shown to, the action counted as a conversion, and the rate below which the hypothesis has failed. Base the bar on the economics, not on hope. If the plan is to acquire customers for ₹1,500 each and a visitor costs ₹15 from the channel you will use, then a page that converts one visitor in a hundred to a paying customer is the break-even line, and a page that converts fewer has to change or the plan does. For a reference point Unbounce reports a median landing-page conversion rate of 6.6 per cent across industries, but that median mixes newsletters and free downloads with demo requests; the right comparison is your own economics.
Then send only the right people. A page shown to friends, investors and a general audience measures their kindness. A page shown through one channel to the one segment you mean to serve, a LinkedIn message to two hundred finance heads, a post in three trade-body groups, a capped Meta campaign to a tightly defined audience, measures the market. Tag the traffic by source so that a good result from one channel and a poor one from another are not averaged into a meaningless middle.
How much traffic the test needs
Here is the part founders skip. A conversion rate measured on a few hundred visitors is a wide range, not a number, and an A/B test that compares two versions needs far more traffic than most early pages will ever get. Evan Miller’s sample size calculator, the tool most growth teams use, shows how fast the requirement grows as the difference you want to detect shrinks. The figure below does the same arithmetic. At a four per cent conversion rate, detecting a thirty per cent improvement, from 4.0 to 5.2 per cent, needs several thousand visitors on each version. At a hundred and fifty visitors a day split between two versions that is more than two months for one test.
A landing page does not ask whether people like the idea. It asks how many of the right people will pay a small price, today, to have it.
Reading the result
Two practical rules come out of the arithmetic. First, at early traffic test big differences, not small ones: a different buyer, a different price, a different promise, not a different button colour. A change that might double the rate, from four per cent to eight, can be detected in five or six hundred visitors per version; a change worth ten per cent needs tens of thousands and at early traffic cannot be detected at all. Second, when there is not enough traffic for a comparison, test one page against the pass mark, and read the result as a range. Four per cent on fourteen hundred visitors means the true rate is probably somewhere between about three and five per cent, which is plenty to decide whether you are near a one per cent bar and useless for deciding between 4.0 and 4.4.
Then talk to the people who converted, and to a few who did not. The page tells you how many; only the conversations tell you why. If the converts are not the segment you aimed at, the result is a new hypothesis about who the customer is, which is worth more than the test you planned. If almost nobody converts, check the order of failures before concluding nobody wants it: was the traffic the right people, did they read past the headline, was the price the obstacle, or the action? Each is a different next test, and the [first channel lesson](/library/finding-first-channel-that-works) is the method for finding traffic that is right.
The two-week test cycle
Run the page in cycles of two weeks. On day one, write the hypothesis, the segment, the action and the pass mark on one line with the date, and launch. On day seven, check that traffic is arriving from the intended source and that the action is being counted; fix plumbing, not copy. On day fourteen, write the visitors, the conversions, the rate and its range, and one sentence: passed, failed or inconclusive. Call five people who converted. Then change one big thing, the buyer, the promise, the price or the action, and start the next cycle. Three cycles in six weeks give a founder more evidence about demand than six months of building on what friends said.
The sample-size arithmetic is standard and approximate; benchmark conversion rates vary widely by offer and traffic. The sources are below. Write the pass mark before you write the headline.