An Option Pool Refills as Employees Leave and Options Lapse

An option pool is the shares a company reserves for employee equity, and it refills whenever someone leaves: unvested options return at once, vested ones when the exercise window closes. Size the option pool on grants net of those returns, and set the window on purpose, because it moves value from the people who leave to those who stay.

The return flow is large. In the years before their IPOs, Twilio, Zoom, Datadog and Snowflake got back between 6.3% and 26.7% of the options they granted each year, about one in seven overall (our arithmetic). Part of it is pay that employees could not afford to keep: on Carta’s data, employees had never exercised 46.1% of the in-the-money options that expired in 2022. That share, and so the size of a reserve drawn from a hiring plan, turns on a three-month clause in the tax code.

Exhibit 1. Eight terms decide what a grant is worth to an employee and how much of the pool it uses

Term	Rule or common setting	What it does to the pool
Share reserve	An ISO plan must state its share count and win stockholder approval; median 12.1% of equity at seed, 16.8% at Series C	Sets the ceiling; raising it takes a new vote
Pricing	The Series A pre-money valuation includes the unallocated pool	Existing holders pay for the reserve
Vesting	25% after one year, then monthly over 36 months	Unvested options return when an employee leaves
Strike	At least fair market value on the grant date	Sets the cash an employee needs to exercise
ISO limit	$100,000 of grant-date value first exercisable per calendar year; the excess is a non-qualified option (NSO)	None; it changes the employee’s tax
Exit window	ISO treatment ends 3 months after employment; 85% of grants allow 90 days	Vested options left unexercised return when it closes
Term	An ISO cannot be exercised more than 10 years after grant	Unexercised options return at expiry
Early exercise	Offered on about 15% of options; an 83(b) election within 30 days	Exercised shares leave the pool; unvested shares bought back return

Source: 26 U.S.C. §§83(b) and 422; National Venture Capital Association (NVCA) model term sheet (2020); Carta (2020, 2023, 2026).

An option pool is a running balance, and before four IPOs one option in seven came back

For incentive stock options, the size of the reserve is a legal requirement. Options qualify as incentive stock options (ISOs) only under a plan that states “the aggregate number of shares which may be issued under options” and that stockholders approve within 12 months of its adoption (26 U.S.C. §422).

On Carta’s data for rounds raised from 2021 through 2025, the median employee pool is 12.1% of a company’s equity at seed; by Series C, at 16.8%, it exceeds the median founding team’s 16.1% (Carta, 2026). Carta’s guide gives about 10% as a rule of thumb (Carta). The NVCA model term sheet prices a Series A on a pre-money valuation that includes the unallocated pool, so the existing holders pay for every share reserved (NVCA, 2020). Benchmarks and term sheets treat the option pool as a stock and ignore what flows back into it.

The plans spell out the inflow. Twilio’s 2016 plan says that shares under awards that are forfeited, cancelled or expire “will be added back to the shares of Class A common stock available for issuance” (Twilio, 2016). Dividing the options forfeited, cancelled or expired in a period by the options granted in it gives the share of each year’s grants that came back.

Exhibit 2. Four software companies got back 6.3% to 26.7% of their grants each year before their IPOs

Company	Period	Options granted	Options returned	Returned as % of granted	Shares available at period end
Twilio	2014	6,844,749	1,825,208	26.7%	not reported
Twilio	2015	6,919,556	1,409,838	20.4%	not reported
Zoom	Year to Jan 2017	5,808,000	628,088	10.8%	5,578,104
Zoom	Year to Jan 2018	9,765,200	646,969	6.6%	10,459,873
Zoom	Year to Jan 2019	9,192,700	580,927	6.3%	1,848,100
Datadog	2017	12,841,968	1,623,264	12.6%	8,939,016
Datadog	2018	11,236,803	2,367,012	21.1%	69,225
Datadog	Jan to Jun 2019	4,261,980	669,387	15.7%	976,632
Snowflake	Year to Jan 2019	25,229,343	1,644,213	6.5%	5,479,974
Snowflake	Year to Jan 2020	46,934,532	7,831,769	16.7%	412,401
All ten periods		139,034,831	19,226,675	13.8%

Source: stock option activity tables in the final prospectuses (Form 424B4) of Twilio (2016), Zoom (2019), Datadog (2019) and Snowflake (2020). Returned options are those each table reports as forfeited, cancelled or expired. Datadog’s 2019 figures are unaudited. Ratios are our arithmetic and compare flows within a period, not cohorts. The four companies were successful and had large stocks of outstanding options; a seed-stage company will get back less.

The balances show how the pools were run. Datadog ended 2018 with 69,225 shares available, about two days of its 2018 granting pace (our arithmetic), and added 4,500,000 in the first half of 2019. Snowflake’s balance fell to 412,401 in January 2020 before the company authorized 20,870,187 more shares (Snowflake, 2020). After an IPO the top-up becomes automatic: Twilio’s 2016 plan and Figma’s 2025 plan each grow by up to 5% of the shares outstanding every 1 January (Figma, 2025). Before it, Datadog, Zoom and Snowflake each ran the option pool as a balance and topped it up when it ran low. A pool sized on gross grants charges existing holders for options that will come back.

But part of the refill is vested pay that leavers could not afford to keep

The inflow has two sources. Unvested options return by design, since a leaver forfeits whatever has not vested. Vested options return only if the leaver does not buy them before the exercise window closes; then, in Carta’s words, they “disappear back into the company’s option pool to be reissued to other employees” (Carta). The filings do not split the two streams; Carta’s data size the second.

Carta’s 2022 report, covering more than 1.5 million US employees with equity, found that 46.1% of in-the-money options that expired in 2022 had never been exercised, and that 85% of grants allowed 90 days after departure (Carta, 2023). In a separate sample of 386 private companies valued above $1 billion, nearly 50,000 workers walked away from fully vested options in 2022, with a net value of more than $1.8 billion at 409A prices. Employees exercised 44% of vested options before they expired in the first quarter of 2022 and 28% a year later (Carta, 2023).

Cost is the reason given most often. Asked why they had not exercised, 23% of the employees in Carta’s survey said they could not afford to and 18% that the risk was too high. Another 13% were “afraid of making a mistake or thought they already owned them”. The survey is self-reported and covers Carta’s users only.

Life insurers run on the same arithmetic. Daniel Gottlieb and Kent Smetters open their study with it: “Most individual life insurance policies lapse, with lapsers cross-subsidizing non-lapsers” (Gottlieb & Smetters, 2021). In their surveys, unexpected liquidity needs explained 15.4% of recent lapses. A pool refilled by expired options works the same way: the vested options that leavers abandon pay for grants to those who stay. The deadline is where the two differ. A policy lapses whenever the premiums stop; an employee has already earned the options and must buy them within about 90 days of leaving, on terms the company set.

Therefore the tax code, more than the hiring plan, sets the terms of each grant

The 90-day window comes from one clause. An ISO keeps its tax treatment only if the holder was an employee from grant until “the day 3 months before the date of such exercise” (§422(a)(2)). Nothing obliges a company to end the option after three months. If it allows longer, options exercised later are, in Carta’s phrase, “taxed like a non-qualified stock option (NSO) grant if exercised after 3 months”, which means income tax on the spread at exercise. Most companies stop at the boundary: in Carta’s unicorn sample, 82% had a median window of 89 to 92 days.

The incentive stock option entered the tax code on 13 August 1981, in Public Law 97-34, with a $100,000 annual cap on grants. The 1986 tax act recast the cap as $100,000 of options “exercisable for the 1st time” in a calendar year, valued at the grant date. The figure has never been indexed: $100,000 in August 1981 is about $363,000 in August 2026 prices (our arithmetic), on the BLS consumer price index.

A one-year cliff makes the limit bind sooner. Take a hypothetical grant of 40,000 options at a $6.00 strike, made on 1 January 2026 on the NVCA schedule. Spread evenly, $60,000 of its $240,000 grant-date value would become exercisable each year. With the cliff, the first calendar year of vesting takes the 25% cliff tranche plus 11 monthly tranches: 47.9% of the grant, or $115,000, so 2,500 options worth $15,000 become NSOs. For a January grant the all-ISO ceiling falls from $400,000 to about $209,000. A rising 409A price pushes later hires over the line first, since the same slice of the company carries a higher grant-date value.

The other way round the window is to exercise early. Where a company allows it, an employee can buy unvested shares at grant and elect under §83(b) to be taxed at once on any excess of their value over the price paid. The election is due “not later than 30 days after the date of such transfer” (26 U.S.C. §83) and “may not be revoked except with the consent of the Secretary”. If the shares are later forfeited, “no deduction shall be allowed in respect of such forfeiture”.

Few employees get the choice. Only about 15% of options on Carta could be exercised early in 2020, a share unchanged since 2010 (Carta, 2020). Where offered, it is used: 11.8% of early-exercise grants showed some exercise after 14.5 months, against 2.2% of other vested grants.

But options retain people while they are unvested, so a short window buys little loyalty

The research suggests options keep and sort people more than they motivate them. Paul Oyer and Scott Schaefer tested three explanations and concluded: “We reject an incentives-based explanation for broad-based stock option plans, and conclude that sorting and retention explanations appear consistent with the data” (Oyer & Schaefer, 2003). Serdar Aldatmaz, Paige Ouimet and Edward Van Wesep found that after a large broad-based grant “employee turnover falls at the granting firm”. But “a large fraction of the reduction in turnover appears to be temporary”, and turnover rises again in the third year (Aldatmaz, Ouimet & Van Wesep, 2018). Both studies cover public companies, whose employees can sell vested shares, so they say less about startups than their samples suggest.

If the pattern holds in private companies, retention works through what has not yet vested, which points to refresh grants rather than the window. A short window adds a different hold: vested options that cost cash to keep. Carta argues that a restrictive window can keep an employee “tethered to a job they would have otherwise left”, producing staff who rest and vest. That is the opposite of the sorting Oyer and Schaefer describe.

Longer windows are spreading: in the first quarter of 2023, companies on Carta offered an extended window on 20% of options for terminated employees. The cost lands on the pool: “Extending PTEP means it will take longer for options to return to the pool”, so some companies will need a larger one. The choice is between pool shares and leavers’ pay, and the default makes it without anyone deciding.

Size the option pool on net grants, then decide the window

A pool built this way has three lines: gross grants from the hiring plan and refreshes, expected returns, and a date for the next top-up. Exhibit 3 applies them to a hypothetical company with 10.0M fully diluted shares after a Series A at a $40M post-money valuation ($4.00 a share), planning 16 hires and refresh grants for 20 employees over the 18 months to its next round.

Exhibit 3. A hypothetical post-Series A pool: 5.2% of the company gross, 3.8% to 4.9% net

Line	Options	% of fully diluted shares	At the Series A price
New-hire grants, 16 hires	420,000	4.20	$1.68M
Refresh grants, 20 employees at 5,000	100,000	1.00	$0.40M
Gross grants	520,000	5.20	$2.08M
Net if 6.3% comes back (Zoom’s lowest year)	487,240	4.87	$1.95M
Net if 13.8% comes back (all ten periods in Exhibit 2)	448,240	4.48	$1.79M
Net if 26.7% comes back (Twilio’s highest year)	381,160	3.81	$1.52M
A 10% template pool	1,000,000	10.00	$4.00M

Hypothetical company and hiring plan: a VP of engineering at 100,000 options, 4 senior engineers at 35,000, 6 engineers at 15,000, a head of sales at 60,000 and 4 other roles at 7,500. Return shares from Exhibit 2. Values at the preferred price, which overstates common stock; computed by us.

Between gross and net sit 0.3 to 1.4 points of the company, $131,000 to $555,000 at the round price. The template is the larger miss: a 10% pool reserves about twice the gross plan, and in the pre-money the existing holders pay for the surplus. A young company, with few options outstanding to come back, should use the low end of the range until its own ledger gives a better figure. So should a company that extends its window, because vested options will return later, if at all. Then it can schedule the top-up, as Datadog and Snowflake did, rather than reserve years of grants.

Employees value options at about half their cost, Hall and Murphy found

Brian Hall and Kevin Murphy made the case against options in 2003. Modelling undiversified, risk-averse employees, they found that “employees value options that have just been granted with an exercise price equal to the market price at only about half of their cost to the firm”, and less still when vesting is long (Hall & Murphy, 2003). Their explanation for why companies grant them anyway: “boards and managers incorrectly perceive stock options to be inexpensive because options create no accounting charge and require no cash outlay” (Journal of Economic Perspectives). The second half of that sentence still describes most startups.

Hall and Murphy’s gap is real, and the 90-day window widens it. An option that must be bought for cash within three months of leaving, or lost, is worth less to a risk-averse employee than one that can wait for an exit. Hall and Murphy studied public companies, whose employees can sell; most startup employees cannot.

What survives is the reason startups grant options at all: equity lets a company hire “without bloating your cash burn rate”, in Carta’s phrase, and it sorts for employees who believe in the company’s future. In John Hand’s study of private venture-backed firms, equity values suffered more when options reached too few levels of staff than too many (Hand, 2004). The remedy is to narrow the gap with longer windows, early exercise and refresh grants, rather than to grant fewer options or reserve more.

A quarterly review of the option pool

The person who runs the equity plan and the directors who approve grants read the same ledger, so one list serves both.

• The running balance: shares available, granted and returned in the quarter, and the month the pool runs out at the current pace.
• The share of grants that came back, from the company’s own option ledger, with the next pool sized on net grants, as in Exhibit 3. A 10% template pool is a forecast to check against the hiring plan and that history.
• The vested options that expired unexercised in the year, and what they were worth at the 409A price.
• The exit window and early exercise, decided in writing as pay decisions and costed in pool shares. The 90-day default follows a tax rule; no law requires it. Before any extension, both sides get a price: the pool shares that will not come back, and the pay that leavers keep.
• Each new grant, checked against the $100,000 ISO limit in its first vesting year: on a one-year cliff, a January grant puts 47.9% of its value into one calendar year.
• The 30-day 83(b) deadline, written into the grant letter of any early-exercise grant.

The pool lives as a balance, and the window decides its refill

Boards set an option pool as a percentage once, and it then lives as a balance: grants drain it, leavers refill it and top-ups restore it. Before four IPOs, about one option in seven came back, and how much of that was vested pay turns on a three-month tax clause most companies adopt by default. Founders should size the pool on net grants and choose the window deliberately, since each vested option that comes back through it was someone’s pay. The return rates here come from four companies that went public with large stocks of options outstanding, so a seed-stage company should plan on getting back less until its own ledger shows otherwise.

Sources and notes

• Aldatmaz, S., Ouimet, P. and Van Wesep, E. D. (2018). The option to quit: the effect of employee stock options on turnover. Journal of Financial Economics 127(1), 136–151. Working-paper version, SSRN, 2014.
• Carta. Option Pool Definition: How to Size Your Employee Option Pool; The Post-Termination Exercise Period (PTEP) for Options Explained (web guides).
• Carta (2020). The current state of employee exercise 2020.
• Carta (2023). Hodge, A. and Walker, P., 2022 Employee Stock Options Report; Dowd, K., Startup employees were less likely to exercise stock options in Q1.
• Carta (2026). Walker, P. and Dowd, K., Founder Ownership Report 2026.
• Datadog, Inc. (2019). Prospectus (Form 424B4), filed 19 September 2019.
• Figma, Inc. (2025). Prospectus (Form 424B4), filed 31 July 2025.
• Gottlieb, D. and Smetters, K. (2021). Lapse-based insurance. American Economic Review 111(8), 2377–2416.
• Hall, B. J. and Murphy, K. J. (2003). The trouble with stock options. Journal of Economic Perspectives 17(3), 49–70; NBER Working Paper 9784.
• Hand, J. R. M. (2004). Give everyone a prize? Employee stock options in private venture-backed firms. SSRN working paper; published in Journal of Business Venturing 23(4), 385–404 (2008).
• National Venture Capital Association (2020). Model Term Sheet for Series A Preferred Stock Financing.
• Oyer, P. and Schaefer, S. (2003). Why do some firms give stock options to all employees? An empirical examination of alternative theories. SSRN working paper; published in Journal of Financial Economics 76(1), 99–133 (2005).
• Snowflake Inc. (2020). Prospectus (Form 424B4), filed 16 September 2020.
• Twilio Inc. (2016). Prospectus (Form 424B4), filed 23 June 2016.
• US Bureau of Labor Statistics. Consumer Price Index for All Urban Consumers, US city average, all items (series CUUR0000SA0).
• 26 U.S.C. §§83 and 422, with statutory history notes, via Cornell’s Legal Information Institute.
• Zoom Video Communications, Inc. (2019). Prospectus (Form 424B4), filed 18 April 2019.

Quotations from Oyer and Schaefer, Aldatmaz et al. and Hand are from the papers’ abstracts (their SSRN working-paper versions), and Hall and Murphy’s estimate is from their NBER working paper. Numbers that carry no citation are computed from the prospectuses, Carta’s data and the price index: return shares divide options returned by options granted in the same period, and the 2026 value of the ISO limit scales $100,000 by the CPI-U for August 1981 and August 2026. The grants in the ISO example and the company in Exhibit 3 are hypothetical. The piece describes US federal tax rules and is not tax advice.