पाठशाला Pathshala · उत्पाद Utpād, The product · Lesson 20 · Build
Launch day is not a strategy
One big morning buys attention once, at the product’s weakest moment. Plan the launch as a series of small releases, each to a wider audience, with a loop between them that fixes what the last one showed.
Pathshala, The Founder Library · 11 October 2026 · 7 min read
Founders plan launch day the way families plan a wedding: one date, every guest, everything polished, and the sense that the result will be decided in a few hours. Products do not work that way. The morning brings a crowd, most of the crowd leaves, and the founder learns little that a smaller crowd would not have taught more cheaply.
This lesson replaces the date with a series. It explains what a launch day buys and what it costs, sets out a ladder of releases from a dozen named users to a public relaunch, describes the loop that should run between each rung, works an example and closes with a six-week calendar a team can run.
What a launch day actually buys
A launch on Product Hunt, Hacker News, LinkedIn or in the press buys one thing: a spike of attention from people who did not know the product existed. That is worth having. It brings sign-ups, a handful of early supporters, sometimes an investor email and a screenshot for the deck. It does not bring retention, because retention depends on the product and the people the spike brings are mostly curious rather than in pain.
The cost is less visible. A big launch spends the product’s first impression on its weakest version, in front of the widest and least forgiving audience, and leaves the team firefighting for a week instead of learning. The cohort it produces is also hard to read. Thousands of tourists dilute the few dozen people who matched the [ideal customer](/library/ideal-customer-profile-on-one-page), so the activation and retention numbers say more about the launch channel than about the product.
Why one big morning is the wrong bet
Paul Graham’s Startup = Growth sets the measure that matters at this stage: a good growth rate during Y Combinator is 5 to 7 per cent a week, 10 per cent is exceptional, and 1 per cent is a sign the founders have not yet figured out what they are doing. Growth rate is a slope, measured week after week. A launch day is a step. A team that judges itself by the step learns nothing about the slope, and the slope is what compounds.
Reforge’s essay on growth loops by Brian Balfour, Casey Winters, Kevin Kwok and Andrew Chen makes the same point with an example: one initiative that produces 500 engaged users once, against one that produces 20 new users in its first week and grows 10 per cent a week. The second is the one they prefer, because its output feeds back into its input. A launch day is the first kind of initiative. A launch series, done properly, is the second.
Even the platform built around launch day does not ask for a single one. Product Hunt’s launch guide recommends 12:01 am Pacific Time for makers who plan ahead, says the best day to launch is the day you are most prepared, and answers the question of how often with: as often as you have significant new product iterations. Google ran Gmail with a beta label for more than five years, as its own announcement of dropping the label in July 2009 noted, and kept widening the audience throughout. A launch is a sequence of moments when a wider group is let in.
The launch ladder
Plan five rungs. Each lets in a wider audience, and each has an exit test that must pass before the next. Rung one, named users: ten to thirty people chosen by hand from the interviews, onboarded on a call. Exit when at least half complete the core action in the first session and come back in week two. Rung two, a community: one community where the ideal customer already gathers, a WhatsApp group of CA firms, a Slack for Indian product managers, a district traders’ association. Exit when self-serve sign-ups activate at the rate the named users did, without a call.
Rung three, a public release: the website goes live, pricing is visible, and the founder posts in their own channels. Exit when first-month retention for this cohort is within a few points of rung two’s. Rung four, the launch moment: Product Hunt, Hacker News, a press story, a partner’s newsletter, timed for when the onboarding can handle strangers. Rung five, relaunches: each significant new version gets its own smaller launch, with the change at the centre of the story.
The order matters more than the dates. A team that reaches rung four with an activation rate it has seen hold for two cohorts will keep more of the spike than a team that arrives there first. A team that never passes rung two has learned something more valuable than any launch would have told it.
The loop between launches
Between rungs run the same short loop, two to four weeks long. Read the cohort: what share [activated in the first session](/library/activation-first-session-that-decides), what share was active after a month, and where the rest stopped. Call five who stopped. Choose the single largest problem and fix it. Ship the fix, then let in the next audience. The loop is what makes a series better than a spike, because each cohort arrives at a product that has learned from the one before.
The figure compares the two plans over the same six months. The single launch brings every sign-up at once, at today’s retention. The series brings fewer people per launch but each loop lifts retention a few points before the next audience arrives.
At the defaults one launch of 3,000 sign-ups at 8 per cent retention leaves 240 active users. Six launches of 500 each, with three points of retention gained per loop, leave about 465, from the same number of sign-ups, and the product ends the half-year retaining 23 per cent instead of 8. Set the lift to zero and the series wins nothing: the advantage comes entirely from fixing things between launches. That is the honest test of whether a team is running a series or just launching six times.
A launch day is a step. Growth is a slope. Spend the launch on learning the slope, one wider audience at a time.
A worked example: an invoicing app for freelancers
A three-person team in Kochi builds an invoicing and GST app for freelance designers and developers. The founders had planned a Product Hunt launch for the week the app was ready. They run the ladder instead. Rung one: 25 freelancers from their interviews, onboarded on calls. Eighteen send an invoice in the first session; eleven send another in week two. The calls reveal that most freelancers bill foreign clients and need export invoices under a letter of undertaking, which the app did not support. Three weeks go into that.
Rung two: a freelancers’ community of a few thousand members on Discord. 340 sign up over a fortnight; 41 per cent send a first invoice without help, close to the named users. First-month retention is 19 per cent, and the cancellations cluster at the payment-reminder step. Two weeks go into automated reminders on WhatsApp. Rung three: the public site, pricing at ₹299 a month, posts by the founders; 900 sign-ups over a month with first-month retention at 24 per cent. Only then rung four, a Product Hunt launch built around the export-invoice feature, which the community had already asked for by name. The launch brings 2,100 sign-ups and the retention holds at 22 per cent, because the product they meet has been through three loops.
The six-week launch calendar
Write the ladder on one page before the first release: the five rungs, the audience for each and its exit test as a number. Then run each rung on the same calendar. Week one: release to the rung’s audience and onboard the first arrivals personally. Week two: read activation and the first-week return, call five people who stopped. Weeks three and four: fix the largest problem and ship it. Week five: read first-month retention for the cohort against the exit test. Week six: pass the test and open the next rung, or run the loop again on the same audience.
Every Monday in the [weekly metrics review](/library/weekly-metrics-review-one-page-one-hour), put weekly growth in active users beside the cohort numbers. Keep the big launch for rung four and spend it on a change worth announcing. After that, give each significant release its own smaller launch, as [shipping weekly](/library/shipping-weekly-cadence-that-compounds) makes routine. A year in, the team should be able to point to six or eight launches, each to a wider audience, and a retention line that rose after every one.
The figure is a simplification and the Kochi company is illustrative; the sources are below.
Sources
- Paul Graham, Startup = Growth, September 2012 — 5–7 per cent a week is good during YC; 10 per cent is exceptional; 1 per cent means the founders have not figured it out.
- Brian Balfour, Casey Winters, Kevin Kwok and Andrew Chen, Growth Loops are the New Funnels, Reforge, July 2018
- Product Hunt, launch guide and FAQ
- Google, Google Apps is out of beta (yes, really), Official Google Blog, 7 July 2009 — Gmail had worn the beta label for more than five years.