The Cap Table: Who Holds What, and What Dilutes It

A cap table lists every holder, the shares each one holds and the percentage those shares produce. The share count is the fact and the percentage is derived from it. The total in issue grew across three rounds at Aravalli Learning Systems Private Limited, an invented company, so the founders' percentage fell from 70.0 to 56.0 to 44.8 while they held the same 70,00,000 shares throughout.

Almost everything people find confusing about private companies is really one confusion about one arithmetic operation. A percentage is a division. Two numbers go into it, and only one of them belongs to the person the percentage describes. Change the other one and the percentage moves without anybody touching anything the person holds. Three rounds at one invented company show the operation happening, and the total in issue is the only number that moves in any of them.

What is a cap table, and which of its two columns is the fact?

Four neighbours buy one second-hand delivery van between them. Somebody writes on a sheet of paper how many parts each of them paid for: eleven parts, seven parts, four parts, three parts. Twenty-five parts in all. The fractions follow from the parts and from the total, so nobody needs to be told what fraction each person has bought. Nobody writes the fractions down either. There is no point in writing them down when the sheet gives them up at any time.

A cap tableThe list of who holds how many shares in a company, with the percentages those shares produce. is that sheet, kept for a company, in shares instead of parts. A cap table is built out of the company's own share registerThe record of who holds how many shares, which is what a cap table is built from.. The register is a company record rather than anything a market publishes. Anything touching a company's own register, its charges and its filings sits with the Ministry of Corporate Affairs at mca.gov.in.

Where the analogy stops is that a company can print new parts. When Aravalli sold new shares, the twenty-five became thirty-one, and every fraction on the sheet moved even though nobody handed over a single part of what they already had. The share count is the fact and the percentage is the arithmetic done on it, so a percentage that moves is not evidence that a share count moved.

The direction of that arrow settles everything that follows. Each holder's issued sharesShares actually in existence and held by somebody. divided by the total in issue gives the percentage. The share counts are the input, the total is the input, and the percentage is the output. Nothing is ever recorded as a percentage anywhere, and a company that stored percentages instead of counts would have no way of adding a new holder without rewriting every row by hand.

Who has a row, and is the reserved pool one of them?

Aravalli Learning Systems Private Limited, invented, ends with seven rows. Two founders, called founder A and founder B here. An employee option pool. Nilgiri Venture Fund I, the invented venture vehicle that came in at the seed. An angel syndicate of six individuals. Then one more row for each priced round that followed: the Series A investor and the Series B investor. Seven rows, and every one of them is a holderAnybody or anything with a row on the register, including a reserved block with no name against it. in the arithmetic even though only six of them are people or institutions.

The odd one is the option poolShares set aside for employees, including ones not yet hired.. The pool is a block of 10,00,000 shares set aside for employees, and at the seed not one of them has been handed to anybody. Some of the people they are meant for have not been interviewed yet. There is no name against the row and there is nobody to ring up about it.

A wedding hall makes the point. One table is laid for guests who have not confirmed. Nobody is sitting at it. The empty table still takes up floor space, it is still counted when the caterer works out the per-head cost, and pretending it is not there gives the wrong number for everybody else. The reserved pool is not a person and it is still a row, and leaving it out of the total makes every other percentage in the table wrong.

Most broken cap tables come from the reserved pool, and the fault is almost never arithmetic. Somebody counts the pool in one round because the spreadsheet they inherited did, and leaves it out in the next because the spreadsheet they built did not, and the two rounds then quietly describe two different companies. The mixed-basis failure is worked below with the numbers attached.

What does the whole table look like across all three rounds?

Everything so far has been one round at a time. All three now come together. Aravalli raised three times, on Aravalli's own clock: the seed in Aravalli Year 1 at Rs 30.00 a share, the Series A in Aravalli Year 3 at Rs 120.00 a share, and the Series B in Aravalli Year 5 at Rs 240.00 a share. One question about the price comes before the table.

Here is the table. All three rounds sit in one frame there, and the table repays reading in two directions, across a row and then down a column.

Read it across a row first. Founder A holds 42,00,000 shares after the seed, 42,00,000 after the Series A and 42,00,000 after the Series B. The option pool holds 10,00,000 at every one of the three moments. Nilgiri Venture Fund I holds 15,00,000 throughout and did not put in another rupee at either later round. The angel syndicate holds 5,00,000 throughout. Not one existing holder's share count falls anywhere in this table, in any round, by any amount.

Now read it down a column. After the seed the percentages are 42.0, 28.0, 10.0, 15.0 and 5.0, adding to 100.0. After the Series A they are 33.6, 22.4, 8.0, 12.0, 4.0 and 20.0, adding to 100.0 again. After the Series B they are 26.88, 17.92, 6.4, 9.6, 3.2, 16.0 and 20.0, and the column still adds to 100.0. Each column is a complete description of the company at that moment, and the three columns are three complete descriptions of the same company at three different moments.

One number changed between the columns, and the number is at the bottom. The total in issue went from 1,00,00,000 shares to 1,25,00,000 to 1,56,25,000. The seed added 20,00,000 new shares at Rs 30.00, being a raise of Rs 6,00,00,000 on a pre-money of Rs 24,00,00,000. The Series A added 25,00,000 new shares at Rs 120.00, being Rs 30,00,00,000 on a pre-money of Rs 1,20,00,00,000. The Series B added 31,25,000 new shares at Rs 240.00, being Rs 75,00,00,000 on a pre-money of Rs 3,00,00,00,000.

Read together, the two directions define dilutionThe fall in a holder's percentage caused by the total share count rising. exactly. Every percentage in the table fell for existing holders. No share count fell for anybody. The falling was done entirely by the denominatorThe total the percentage is taken out of, which is the number that actually moves..

Why does every existing holder move by the same multiplier?

Four holders moved between the seed and the Series A. The founders went from 70.0 per cent to 56.0. The pool went from 10.0 to 8.0. Nilgiri Venture Fund I went from 15.0 to 12.0. The angel syndicate went from 5.0 to 4.0. Each new figure divided by its old one gives 0.8 every single time, from four holders who have nothing in common and never spoke to each other about it.

The shared 0.8 is not a coincidence and not a property of venture capital. The ratio falls straight out of the arithmetic. Each holder's percentage is its share count divided by the total. The share count did not move, so the only thing that could have changed the percentage is the total, and the total is the same for everybody. The old total divided by the new total is 1,00,00,000 over 1,25,00,000, or 0.8. Every existing holder divides by the same number, and that one ratio applies to all of them at once.

Here is the everyday version. One plot of land is held between four cousins in equal shares, so each has a quarter. A fifth cousin is added to the deed on the same footing. Nobody surrendered anything, no boundary was moved, and every one of the original four now has a fifth instead of a quarter. The multiplier is four fifths, it is 0.8, and it applied to all four of them at the same instant because the only thing that moved was the count of names.

Dilution is a denominator effect, and that is why one multiplier explains an entire column of a cap table. The whole column is one multiplication, so a dilution table never has to be computed row by row.

What happens when two rounds of the same multiplier compound?

Aravalli's three rounds have an unusual property, worth stating plainly before it is used. Each of the two later rounds sold exactly one fifth of the company. The Series A investor took 25,00,000 shares of a post-round 1,25,00,000, or 20.0 per cent. The Series B investor took 31,25,000 of 1,56,25,000, or 20.0 per cent as well. The matching fifths are a consequence of the numbers chosen for this invented company, not a claim about how rounds usually go, and rounds in the world sell whatever the two sides agree on.

But because it is true here, the arithmetic becomes unusually clean. Every existing holder is multiplied by 0.8 at the Series A and by 0.8 again at the Series B. Two multiplications, not two subtractions, and 0.8 times 0.8 is 0.64. So every holder present at the seed ends the Series B with 64.0 per cent of the percentage it began with.

Check it against the table. The founders were 70.0 per cent; 70.0 times 0.64 is 44.8, and the table says 44.8. Nilgiri Venture Fund I was 15.0; 15.0 times 0.64 is 9.6, and the table says 9.6. The pool was 10.0; 10.0 times 0.64 is 6.4. The angel syndicate was 5.0; 5.0 times 0.64 is 3.2. Four holders, one multiplier, no rounding anywhere.

The intuition this defeats is subtraction, and it defeats it badly. Sell 20.0 per cent and then sell 20.0 per cent again, and it feels as though 40.0 points have gone, taking the founders from 70.0 to 30.0. The gap between 30.0 and 44.8 is not a rounding difference; it is 14.8 points of a company, and it exists because the second round's fifth was taken out of a company the founders already held less of. Rounds compound multiplicatively, so a reader who subtracts will be wrong by more with every additional round.

What are the two halves of dilution, and why must they be read together?

Dilution has two halves, and both have to be read at the same moment. Reading one half and stopping is exactly the mistake. The founders' holding fell from 70.0 per cent after the seed to 56.0 after the Series A to 44.8 after the Series B, a fall of 25.2 points, and they kept 64.0 per cent of the percentage they started with. And over the same three rounds the same unchanged 70,00,000 shares were priced at Rs 21,00,00,000 at the seed price of Rs 30.00, at Rs 84,00,00,000 at the Series A price of Rs 120.00, and at Rs 1,68,00,00,000 at the Series B price of Rs 240.00. The price a share rose from Rs 30.00 to Rs 240.00, exactly eight times, so the rupee figure on an unchanged share count rose by exactly eight times too. Both of those are true at the same instant, of the same 70,00,000 shares, and an account that gives only one of them has taught half of dilution.

The second half needs one more fact standing beside it. Neither figure is a valuation of anything. Rs 1,68,00,00,000 is 70,00,000 shares multiplied by Rs 240.00, and Rs 240.00 is the price one buyer paid for newly issued shares on one occasion. Nobody has offered the founders that price, or any price, for a single one of their shares. The same arithmetic run at the seed gives Rs 21,00,00,000, and that was not a valuation either. A share count multiplied by a round price is a share count multiplied by a round price, and calling it anything grander is where a lot of nonsense about private companies begins.

The same care applies to the word post-moneyThe share count after a round multiplied by that round's price.. Post-money is the same operation done on the whole register rather than on one row. Aravalli's post-money figures are Rs 30,00,00,000, Rs 1,50,00,00,000 and Rs 3,75,00,00,000, and each is simply the total in issue multiplied by that round's price. Post-money is a useful label for the round, and it is not what anybody would pay for the company. The table has never known what anybody would pay.

Two of the numbers above are unusually clean, and cleanliness invites a wrong conclusion. The 0.64 and the exact eight are consequences of the figures chosen for this invented company and nothing more. Rounds in the world sell whatever the two sides agree on, and there is no reason at all for a price ratio across three rounds to land on a whole number. Use the two identities as a way of seeing how the arithmetic works, never as a pattern to expect somewhere else.

Move it back to the seed and forward again a few times and the shape of the thing becomes hard to unsee. The upper bar shortens, the lower bar lengthens, and the strip between them, the one carrying the share count, never moves at all. The strip carrying the share count is the fact. The two bars are two different calculations performed on it, using two different second numbers: the total in issue for one and the round price for the other.

Is this table counting the reserved pool, or leaving it out?

There are two totals to divide by, and the difference between them is the pool. Counting only shares that exist and are held by somebody gives the issued total. Counting as though every reserved share were already handed out gives the fully dilutedCounted as if every reserved share were already issued. total. Aravalli's table takes the second basis and takes it in every round: the pool's 10,00,000 shares sit inside 1,00,00,000, inside 1,25,00,000 and inside 1,56,25,000.

Neither basis is the correct one in some absolute sense. Both are used, both appear in real documents, and a table that says which one it is using has done its job. Switching between the two bases from one round to the next moves the denominator for reasons that have nothing to do with the company, and switching like that is never acceptable.

Work the damage through with Aravalli's own numbers, derived here rather than taken from the locked record. Suppose the seed column were computed on 1,00,00,000 with the pool counted, and the Series A column on 1,15,00,000 with the pool left out. The founders' 70,00,000 shares would read 70.0 per cent and then 60.9 per cent. Consistently counted they read 70.0 and then 56.0. The two answers are 4.9 points apart, and not one share moved between the two versions.

Worse, the fault hides. The implied multiplier in the mixed version is 1,00,00,000 over 1,15,00,000, or 0.870, and 0.870 is a perfectly plausible looking number. An analyst who expected 0.800 and found 0.870 will go hunting for a share issue that never happened, and may well find something else to blame. The error is not in any row. The fault is in which total the rows were divided by.

What is a cap table silent about?

A cap table is a narrow document and most of its usefulness comes from being narrow. People read things into a cap table constantly, and what it does not carry is worth setting out.

A cap table says nothing about cash. Aravalli's table does not say whether the Rs 75,00,00,000 raised at the Series B is still in the bank, has been spent on salaries, or was never fully drawn. A company can have a magnificent register and no money.

A cap table says nothing about control. Percentages and votes are two different things, and rights to appoint a director, to block a decision or to be paid first live in the company's own documents, not in the share count column. A holder at 9.6 per cent may sit on the board and a holder at 20.0 per cent may not.

A cap table says nothing about what anything is worth. There is no price on the table at all except the price of the last roundOne occasion on which a company issues new shares at an agreed price., and that price was struck between two parties on one day for newly issued shares. The price does not carry over to a block somebody wants to sell, or to the whole company, or to the same shares next week.

A cap table answers exactly one question: who holds how many shares out of how many. Every other question brought to it is answered by a document somewhere else. The narrowness is not a weakness. A record that tried to carry cash, control, rights and price as well would be wrong about all four within a month.

Which two checks catch a broken cap table before anybody relies on it?

Both checks are done inside the table, with nothing brought in from outside it, and together they take under a minute.

The first check is that each round's percentage column adds to 100.0. If it does not, the total in issue is wrong somewhere, and the commonest single cause is a reserved block counted in one round and left out of the next. Do not start by hunting through rows. The total is the number every row depends on, so start there.

The second check is the multiplier. Every holder that existed before the round has its new percentage divided by its old one, and the same number should come out for all of them. In Aravalli's Series B column, 44.8 divided by 56.0 is 0.8, 9.6 divided by 12.0 is 0.8, 6.4 divided by 8.0 is 0.8, 3.2 divided by 4.0 is 0.8, and 16.0 divided by 20.0 is 0.8. Five holders, one number, and the check passes.

If one holder's ratio differs from the rest, that holder's share count moved and somebody should be able to say why; if every ratio differs from the expected total ratio, the denominator moved instead. The two checks between them locate the fault before a single row has been read in detail, and that is why they are worth doing first rather than last.

How does a lender, an analyst or a company secretary actually use this?