विमर्श Vimarsh · Udyamnīti

The Real Startup

Almost nothing is coming to kill your venture. The first of the countdown to NS Transform 2027.

Nikhil Sharma · उद्यमनीति Udyamnīti · 30 September 2026 · 17 min read

A company that was winning

In 2019 I helped build a consumer e-commerce business that was profitable in its first month.

That is rare enough that I should say it twice. Not profitable in year three after a raise. Profitable in month one. We took it through search and shopping campaigns and within about eighteen months it sat in the top three in its category in India, holding that position against Amazon and Flipkart, which are not companies you beat by accident.

Inside three years it was gone.

Amazon did not kill it. The market did not turn. Demand was there the whole way down. It died in a room, between people who had once agreed about everything and had never once written any of it down. Access was revoked. Logins changed. Lawyers appeared. The last eighteen months of that company were spent entirely on its own internal weather while the category it had led went to somebody else.

I have thought about that company for years, and the conclusion I have reached is the least fashionable thing I could say to an Indian founder.

Almost nothing is coming to kill you. You are the risk.

Not for want of advice

No Indian founder has ever died of too little advice. There is more of it than there has ever been. Podcasts. Accelerators. Twelve-week programmes. Men on LinkedIn who have never made a payroll explaining payroll. A founder in Gurugram in 2026 can get a plausible answer to any question inside an hour and most of the answers are not even wrong.

The company I began with had the advice. It had the model. It had the customers. What it did not have was anything that happened when a decision taken on Monday had not been acted on by Friday.

That is where a venture dies. Not in the strategy. In the gap between deciding and doing. The decision is taken. Everyone nods. A week passes. The thing has not moved and nobody says so because nothing was ever going to happen if nobody said so. Then another week. Then the quarter.

Peter Drucker wrote this down in 1967 and nobody has improved on it since.

Good intentions. That is the correct name for most of what happens in an Indian startup's Monday meeting. Steve Blank put the other half of it in a line he has been repeating since 2010 and which founders quote without hearing. No business plan survives first contact with customers. He borrowed the shape of it from Moltke, who said it about armies, and it is truer of companies than it ever was of armies because a company meets its enemy every day.

Drucker and Blank are describing the same gap from opposite ends. Drucker is saying the decision is worthless until it has become somebody's work. Blank is saying the plan is worthless until the market has corrected it. Between those two points — the plan you wrote and the work you actually did — sits the whole of the company. Everything else is a deck.

A startup does not fail for want of advice. It fails in the gap between deciding and doing.

The number everybody quotes and nobody checks

Ninety percent of startups fail. You have heard it. It is on every panel and in every deck. Chase it down and the trail ends at one line in Startup Genome's Global Startup Ecosystem Report for 2019: building a successful business is every entrepreneur's goal, but only one in twelve succeed in doing so. The report rests that on a dataset of about 34,000 companies, does not say how they were sampled or where they sat, and never defines what succeed means. Eleven in twelve not succeeding is not ninety percent failing, and a number with no definition of failure behind it cannot tell you what killed anyone.

Set it aside. What can be sourced is below. I have used the figures the ecosystem itself uses and I have not rounded any of them past what the source states.

Read those four together and the shape is plain. The register has never been longer. The government recognised 55,200 startups in the year to March 2026 and the total passed 2.23 lakh. In the same two years the graveyard grew faster than it ever has. 15,921 companies shut in 2023 and 12,717 in 2024 by Tracxn's count, against about 2,300 across the whole of 2019 to 2022. Funding fell from $37.2 billion in 2021 to $10.7 billion in 2023 and has stayed near there since. Two unicorns were made in 2023. The year before it was twenty-three.

The sector read those numbers as a funding story. The money went away so the companies went away. I read them differently and I will say why.

A funding winter does not kill a company that is doing what it said it would do. It kills a company that was being paid to not notice the gap. When capital is cheap the distance between deciding and doing is invisible because somebody else is covering the cost of it every month. When capital is dear the same distance shows up as a hole in the bank balance with a date on it. The winter did not create the failure. It removed the subsidy that was hiding it.

Here is what twenty years of actually standing in these rooms has taught me, which is a narrow sample and a real one.

Of the ventures I have watched die, very few were beaten. A handful ran out of road because the thing simply did not work, and those are the honourable deaths — the founders usually knew by month nine and were brave enough to say so.

The rest died of each other.

Not of competition. Not of funding winters, which are real and which good businesses survive. They died of unwritten agreements, undefined roles, equity handed over in a good mood, and the specific Indian habit of treating a business relationship as a family relationship right up until the moment money arrives, at which point it becomes a family dispute, which is the worst kind there is.

A competitor can only take your customers. A co-founder can take your company.

Our founding epic is a cap table dispute

We should not pretend this is a modern problem or an Indian failing. It is older than both.

The Mahābhārata is, among everything else it is, a property dispute between cousins. Who holds what share. Who was promised what by whom. Whether an assurance given in one generation binds the next. The evidence for reading it that way is not the dice game. It is the Udyoga Parva, the book of effort, in which the two sides spend a whole volume trying to settle out of court and fail.

Yudhiṣṭhira's last offer, carried by Sañjaya, was five villages. In Ganguli's translation: Give us even Kusasthala, Vrikasthala, Makandi, Varanavata, and for the fifth any other that thou likest. Even this will end the quarrel. Then Kṛṣṇa himself went to Hastinapura as envoy and was refused. Duryodhana's answer is the line the whole negotiation is remembered by: even that much of our land which may be covered by the point of a sharp needle shall not be given by us unto the Pandavas. The embassy failed and eighteen days of war followed over a settlement that had been on the table at five villages.

Read it as a founder and the lesson is not that the Kauravas were wicked. It is that nothing was written while the cousins still ate together. By the time the best negotiator in the epic walked into the room, any deal had become a humiliation for whoever accepted it.

We treat it as scripture. It is also the most complete case study in founder conflict ever written, and we have been handing it to our children for two thousand years without anyone drawing the obvious operational conclusion.

Write it down. Write it down while everyone still likes each other.

What the Dutch discovered in 1602, and what broke in 1609

The modern company begins with the Vereenigde Oostindische Compagnie in 1602 — the Dutch East India Company, the first to issue shares that anybody could buy and anybody could sell on.

It was a genuine invention. For the first time ownership separated from management. The man with the capital no longer had to be the man on the ship.

The problem announced itself in seven years.

Isaac Le Maire had been one of the company's largest early shareholders and had fallen out with its directors. In 1609 he organised what is generally recorded as the first shareholder revolt and the first bear raid in financial history — selling shares he did not yet own, agitating against the board, petitioning the government, writing what may be the first shareholder grievance letter in the European record. The directors responded by changing the rules.

Seven years. That is how long it took humanity to invent the joint-stock company and then invent the fight inside it.

The lesson is not that Le Maire was a villain or a hero. Historians still argue. The lesson is structural and it has never changed. The instrument that lets strangers build something together is the same instrument they will use against each other when they disagree. Shares are not a reward. They are a weapon you are handing someone, in advance, in the hope that they never need to use it.

Most Indian founders hand them out as though they were a compliment.

The three documents nobody writes until it is too late

I am going to be boring for four paragraphs because boredom is where the money is.

Vesting. Equity that arrives over four years with a one-year cliff, for everyone, including the founders, including you. The objection is always the same — we trust each other, it feels insulting. The entire point is that it is written when you do trust each other. Vesting is not a prediction that someone will leave. It is a decision, made while everyone is reasonable, about what happens if they do.

Roles with edges. Not titles. Decision rights. Who signs. Who hires. Who can spend what without asking. Who breaks a deadlock, and by what mechanism, when two people who respect each other genuinely disagree and both are partly right. Most co-founder disputes are not about ethics. They are two correct people with no agreed way to be correct at the same time.

Exit before entry. What happens if one of us wants out in year two. What the shares are worth, who may buy them, what the process is. Discuss it in month one when it is abstract and cheap. In year two it is personal and it costs you the business.

None of this is clever. All of it is subtractive — it removes future arguments rather than adding present capability. Which is why it never feels urgent, and why the people who skip it are usually the ones with the most promising companies. Mediocre ventures die before the governance matters. Good ones live long enough to be killed by it.

Why I never sit in the operator's chair

I back founders, and the deal I offer is equity rather than retainers. I bring distribution, a method and leverage. They bring the business and they run it.

And I hold to one rule that costs me money and that I will not break.

I never sit in the operator's chair.

Kautilya would have understood why. The Arthaśāstra is relentless on the point that a king who does everything himself has no state, only a very busy man. Power that is not delegated is not power — it is labour with a title on it. An investor who starts running the company has not rescued it. He has removed the only person who could have.

There is a selfish version of this argument and an honest one. The selfish version is that my time does not scale. The honest one is that a founder who is being operated by someone else stops being a founder within about two quarters, and what you are left with is an employee who owns equity and resents you for it.

So I will argue. I will show the numbers. I will say plainly that I think a decision is wrong. And then they decide, because it is theirs.

That is the last thing the Rigveda says before it stops. Not a prayer for victory or for wealth. A plea for a group of people to be able to agree. Three thousand years later it is still the binding constraint on every venture I have ever seen.

The four beats

I run one method and it is four words.

Commit. Act. Account. Consequence.

What did you say you would do. Did you do it. Who saw. And — the beat everybody skips — what happened because you did not.

Most companies run the first three. Commitments get made in a meeting. Work gets done. Updates get given. And then nothing attaches to the gap, so the gap widens, and within a year the weekly review is a theatre where people describe a company that does not exist.

Consequence is not punishment. It can be as small as the thing being visible to one other person who matters. But it must exist, and it must be agreed in advance, because a consequence invented after the failure is just a fight.

Apply that to founders and it becomes uncomfortable immediately, which is how you know it is the right instrument. What happens if a co-founder misses three quarters. Nobody wants to answer that in month one. Everybody has to answer it in year three, at a much worse price, usually through lawyers.

You are not choosing whether to have the conversation. You are choosing when, and how much it costs.

The Gītā has a line for the second beat. Founders mostly read past it on the way to the famous one about fruit.

Niyataṃ karma. Not inspired work. Not heroic work. The work that was set. The verse is not asking Arjuna to be brilliant. It is telling him that the body itself does not run on intention and neither does anything built by one. That is the Act beat in eleven words and it was written down a long time before Drucker got to it.

What the crore is for

Kautilya opens the chapter of the Arthaśāstra on embezzlement with a sentence that has nothing to do with embezzlement.

It is the least quoted line in the book and the one most founders need. Everything else — strategy, hiring, product, the Monday meeting — sits downstream of whether the treasury can carry the business across the gap between deciding and doing. Most Indian founders cannot. They decide correctly and then spend the quarter raising the money to act on the decision and by the time it arrives the decision is stale and the market has moved.

So this is the deal I make and I will state it in one sentence because it is the only commercial sentence in this essay. NS Transform carries up to ₹1 crore of operating cost for a founder in return for equity. Not a retainer. Not a loan. I am not going to list what sits inside that figure because every company spends its gap differently and a list would just be another deck. The crore buys one thing. It buys the distance between the decision and the doing, so that the founder can be judged on the four beats rather than on the fundraise.

In return the founder accepts the loop. All four beats. Agreed in advance and written down. That is the price and it is higher than the equity.

The treasury does not make the decision. It is what lets the decision survive the week.

The part that is not about documents

I have made this sound like paperwork and it is not.

That company I began with was built by people who were good at what they did and who, for a while, genuinely liked each other. Nobody set out to destroy it. Everybody involved believed, right to the end, that they were the reasonable one. That is the part nobody writes down afterwards, and it is the only part that matters.

The documents are not there because you expect betrayal. They are there because you will both, eventually, be tired and frightened and convinced you are owed something, and on that day you will want to be governed by a decision made by better versions of yourselves.

Write it down while you still like each other.

That is the whole essay. The rest is just evidence that it has gone wrong before.

The real startup

There is a reason this piece is called what it is.

We use the word startup for the launch. The registration, the logo, the first post, the day the site goes live. That is the easy part and it is not the start of anything. Plenty of companies have done all of it and never actually started.

The real startup is the agreement. The hour where three people who like each other decide, in writing, what happens on the day they do not. Everything that lasts is built on that hour and almost nobody spends it.

So this essay begins a countdown.

NS Transform opens for 2027. A small number of founders. Equity rather than retainers, because I would rather be wrong alongside you than invoice you while you fail. Up to ₹1 crore of operating cost carried from my side, with distribution, method and leverage. The business and the operator's chair stay yours — I have already explained why I will not take it.

Between now and the intake I will publish the rest of this argument in public, piece by piece, so that anyone who applies already knows exactly how I think and can decide they disagree before either of us wastes a meeting.

The countdown starts here.

The wager

This essay, like the long ones before it, ends with something that can be checked rather than admired.

I am saying that the next decade of Indian startup failure will be a governance story, not a funding story — and that the sector will keep reporting it as a funding story because governance failures are embarrassing and capital cycles are not.

Four markers. All public.

1. Founder-dispute disclosures in Indian startups rise as a stated cause of shutdown, while "could not raise" falls as a share. Score it on Inc42's annual list of startups that shut down, which states a reason for each closure, with Tracxn's annual shutdown count as the denominator. 2. At least three Indian companies valued above a billion dollars suffer a public founder or board rupture that materially damages the business, independent of market conditions. 3. Vesting with a cliff becomes standard for founders, not only employees, in Indian seed documentation — visible in the standard templates the ecosystem actually uses. 4. Audit and governance diligence moves earlier, from Series B to seed, as a routine condition rather than an exception.

My call, first published 30 September 2026.

By 2031 I expect three of these four. If fewer than two have happened I was wrong about the mechanism and I will say so here, at the same length.

And because 2031 is a long way off I will add one marker that falls due sooner, so nobody has to take the rest on trust.

By 31 March 2029 the DPIIT register will have passed four lakh recognised startups, and Tracxn will have recorded more than ten thousand shutdowns in each of 2027 and 2028. The register and the graveyard will keep growing together, through a funding recovery, because the thing killing the companies was never the funding. If shutdowns fall below ten thousand in either year while the register keeps growing, the funding story was right and mine was wrong, and I will say so here.

The real startup is not the launch. It is the hour you agree what happens on the day you disagree.

Nikhil Sharma, writing as Nishkrant Nikhil. Gurugram, 30 September 2026. The first of the countdown to NS Transform 2027.

Sources: DPIIT recognition figures to 31 March 2026 as reported by Inc42 from the Ministry of Commerce and Industry. Shutdowns in 2023 and 2024 from Tracxn data via Financial Express and The Wire, 24 April 2025. Funding 2021 to 2024 and unicorns minted in 2024 against 2023 from Tracxn's annual report via Business Standard, 19 December 2024; 2025 funding from the same report a year on. Two unicorns in 2023 against twenty-three in 2022 from Tracxn via Outlook Start-up; the all-time count of 110 from Inc42's unicorn tracker. Drucker from The Effective Executive, 1967. Blank from his own site, 8 April 2010, adapting Moltke. Kautilya in Shamasastry's translation, Book II chapter 8. Startup Genome, Global Startup Ecosystem Report 2019, page 20, for the one-in-twelve line and the 34,000-company dataset. The five villages from Udyoga Parva section 31 and the needle's point from section 127, both in K. M. Ganguli's translation. Inc42's annual shutdown list, 23 December 2025, as the scoring source for the first marker. Gītā 3.8 and Ṛgveda 10.191.2 from the standard texts.