पाठशाला Pathshala · वृद्धि Vṛddhi, Growth · Lesson 01 · Build

Growth rate is the only number that matters early

At seed, revenue is small by definition. The rate at which it compounds is the one figure that predicts what the company becomes, and the one figure a founder can manage week by week.

Pathshala, The Founder Library · 10 October 2026 · 9 min read

A company doing ₹2 lakh a month is not interesting to anyone, including its founders. A company doing ₹2 lakh a month and growing seven per cent a week will do ₹67 lakh a month in a year, and that is interesting to everyone. The amount is the same. The rate is the whole difference, and early on it is the only number worth managing.

This lesson is about one idea from one essay and what to do with it on a Monday. It covers where the benchmark bands came from, why compounding makes the rate matter more than the figure, how to choose the one metric to grow, and the point at which a weekly rate stops telling you anything true.

Startup = Growth

In September 2012 Paul Graham published Startup = Growth, and the argument is in the title. A startup is not a small version of a big company and not a company that uses technology or takes venture money. It is a company designed to grow fast. Everything else about startups follows from that one property, and a founder who is not growing fast is running a business, which is honourable, but a different thing.

Graham then gave the numbers that became the industry’s yardstick, drawn from what Y Combinator saw across its companies. A good growth rate during the programme is five to seven per cent a week. A company hitting ten per cent a week is doing exceptionally well. A company managing only one per cent is showing a sign that it has not yet figured out what it is doing. Y Combinator measures weekly partly because there is so little time before Demo Day, but the deeper reason is that a week is the shortest period over which a founder can change something and see the result.

He was also explicit about what to measure. The best thing to measure the growth rate of is revenue. The next best, for companies not charging at first, is active users. Not sign-ups, not downloads, not app installs, not followers. Money, or failing that, people who came back.

Why the rate beats the amount

Graham’s arithmetic is the part most founders remember: a company growing at one per cent a week grows 1.7 times in a year, while a company growing at five per cent a week grows 12.6 times. Take it further. Seven per cent a week is 34 times in a year. Ten per cent a week is 142 times. The difference between one per cent and ten per cent a week is not ten times. Over a year it is eighty-five times.

This is why an investor looking at a seed company spends almost no time on the revenue figure and almost all of it on the slope. ₹2 lakh a month at five per cent a week is ₹25 lakh a month in a year, a run rate of ₹3 crore, and a credible Series A conversation. ₹20 lakh a month at one per cent a week is ₹34 lakh a month in a year and the same conversation a year later, if at all. The second company is ten times bigger today and will be smaller in eighteen months.

It also explains why Y Combinator tells its companies to treat the rate as the thing being managed, not the number. A target of ₹5 lakh a month by December is a target. A target of five per cent a week is a system: every Monday it tells you how much more than last week you must do, the figure rises as you succeed, and missing it for two weeks in a row is a signal you cannot ignore. Sam Altman’s Startup Playbook puts the mechanism in one line: the company does what the CEO measures.

The green line is your company; the dotted lines are Graham’s bands. Set revenue to what you did last month and growth to what you actually managed over the past four weeks, not what you hope for. The readout on the right converts the weekly rate into a monthly one, because Indian subscription businesses invoice monthly and the two must reconcile: five per cent a week is about 23 per cent a month, seven per cent a week is 34 per cent, ten per cent a week is 51 per cent. Then read the number of weeks to double. At five per cent a week it is fourteen. If the company has not doubled since the quarter began, it is not growing at five per cent, whatever the deck says.

Choosing the one metric

The discipline only works if there is one metric. Altman’s playbook says it is valuable to have a single metric the company optimises, and Hariharan told Startup School in 2019 that at the early stage only three or four metrics matter, of which one leads. The reason is not simplicity for its own sake. It is that two metrics can be traded against each other, and a team with two growth metrics will grow whichever is easier that week.

The choice follows the business model. If customers pay, the metric is revenue: monthly recurring revenue for a subscription, net revenue for a marketplace, which means the take rate and never the transaction value. If the product is free while you find the model, the metric is active users, defined as people who did the core action in the period, not people who opened the app. For a two-sided marketplace the metric is usually completed transactions, because they prove both sides showed up. For a B2B product selling into companies, Altman’s advice is to track revenue growth per month, since weekly figures are distorted by a handful of large contracts.

Three tests for the metric you have chosen. It should be a number that goes up only when a customer got value, so sign-ups fail and paid invoices pass. It should be hard to inflate with money, so GMV bought with cashback fails. And it should be the number you would be embarrassed to have flat, which is a surprisingly reliable test.

Hariharan added one more rule that Indian founders routinely break: report the compounded monthly growth rate, not an average of monthly percentages. Grow 50 per cent one month and zero the next and the average says 25 per cent a month while the truth is 22 per cent. Over six months the gap between the two methods can double the implied rate. The formula is end over start, raised to one over the number of months, minus one; it takes a cell in a sheet, and an investor will compute it anyway.

The company is not the number. The company is the slope.

What a growth rate is for

A rate of five to ten per cent a week does not last and is not supposed to. Graham’s point about the YC bands was never that a company should grow like that for years. It was that in the months when a company is searching for its engine, the weekly rate is the fastest available signal of whether the search is working. A week in which the rate rose tells you the change you made last week was good. A week in which it fell tells you to undo it. Over a quarter, that feedback loop finds product-market fit faster than any amount of planning.

It also sets the tempo of the company. A team held to a weekly growth target ships weekly, calls customers weekly and kills features weekly. A team held to a quarterly target meets quarterly. Altman’s version is that growth and momentum are the keys to great execution, and the prime directive is never to lose momentum. The weekly rate is how momentum is made visible.

When weekly growth stops meaning anything

There are three points at which the weekly rate stops being the right instrument, and knowing them matters as much as knowing the bands.

When the base is too small. Going from ₹20,000 to ₹21,000 in a week is five per cent and means nothing; one customer did it. Altman notes the mirror case, where growth is bad in absolute numbers even though it is good on a percentage basis. Below a few lakh a month of revenue, or a few hundred active users, read the weekly rate for direction only and do not quote it to anyone.

When contracts are long and few. A company selling ₹10 lakh annual contracts to enterprises closes two in one week and none for the next six, and its weekly growth rate is a sawtooth. Measure monthly from the start, and once the pipeline is real, quarterly. The rate is still the thing being managed; the period simply matches the sales cycle.

When the company is large enough that annual bands apply. Once a company passes roughly ₹8 to ₹10 crore of annual recurring revenue, which is where the global benchmarks begin, the yardstick changes from weekly to annual. Bessemer Venture Partners’ State of the Cloud 2019 framed it as good, better and best: good cloud companies reach $10 million of ARR in four years and $100 million in ten, better ones in three and seven, and the best in two and five. Their later benchmarks by scale put the top performers at 230 per cent or more annual growth between $1 million and $10 million of ARR, 135 per cent or more between $10 million and $25 million, and 80 per cent or more above $50 million. Neeraj Agrawal of Battery Ventures compressed the same curve into a mantra in The SaaS Adventure in 2015: reach $2 million of ARR, then triple, triple, double, double, double. Those are dollar benchmarks for American software companies; the point for an Indian founder is the shape, which is that expected growth falls as scale rises, and a company still quoting weekly rates at ₹50 crore of revenue is using the wrong ruler.

A fourth, quieter case: when the weekly number is going up because of spend that will not continue. A festive-season campaign that lifts revenue 15 per cent a week for six weeks is a campaign, not a growth rate. Report it, enjoy it, and do not let it into the compounded figure you show investors without saying what caused it.

A weekly ritual, in fifteen minutes

Every Monday morning, before anything else, one founder updates a single sheet with last week’s value of the one metric. The sheet computes the week-on-week rate, the four-week compounded rate and the weeks-to-double. Read the four-week figure, not the one-week figure; the one-week figure is noise. Say out loud which band the company is in: under one, one to five, five to ten, over ten. Then name the one change made last week that most likely moved the rate, and the one change to make this week. Write both down. In thirteen weeks the sheet is the honest history of the quarter and the explanation of the curve, and the explanation is what an investor is actually buying.

If the four-week rate has been under two per cent a week for a month and the company is under a year old, the ritual has done its job: it is telling you the engine is not found, and the work this week is customers, not features. Graham’s one per cent is not a verdict on the company. It is an instruction to go back to the search.

The bands here are Y Combinator’s observations about its own companies in 2012, stated as such. They are the most useful yardstick in circulation, not a law. Read Graham’s essay in full; it takes twenty minutes and will save a year.

Sources

1. Paul Graham, Startup = Growth, September 2012 — The weekly bands, what to measure, and the 1.7× versus 12.6× arithmetic.
2. Sam Altman, Startup Playbook, Y Combinator — One metric, the company does what the CEO measures, never lose momentum.
3. Y Combinator, Startup School Week 3 Recap: Anu Hariharan and Adora Cheung, September 2019 — Compounded monthly growth rate, not averages; only three or four metrics matter early.
4. Bessemer Venture Partners, State of the Cloud 2019 — The good, better, best framework for time to $10 million and $100 million of ARR.
5. Bessemer Venture Partners, Scaling to $100 million — Annual growth benchmarks by ARR scale.
6. Neeraj Agrawal, The SaaS Adventure, TechCrunch, February 2015 — Triple, triple, double, double, double.