The Employer Pay-Calendar Reference: Vest Shape, Bonus Date, Clawback

I keep this open while I build a source-company sheet or price a buyout. It has one row per pay mechanic, and each row says where to verify the employer's value publicly, what confidence grade that value earns, what move window or buyout cost follows from it, and how it most often misleads. It is for in-house recruiters, sourcers, talent leaders, and founders who need to know when a company's people can leave cheaply and how much forfeited value an offer has to replace.

How do I grade a pay-calendar value?

Grade every value A, B, or C. A means a filing states that exact value. B means two or more secondary sources agree. C means a single source, or sources that conflict. The grade stays attached to the value all the way into the outreach window and the buyout, so a C on any input makes the output a C.

The scale is my own proposal. Nobody has published a standard for it. It works because public filings split the answer across documents:

• The 10-K share-based compensation note reports the service period, because that is what the accounting expense calculation needs. It rarely gives the yearly split or the dates.
• Form 4 filings show actual vest dates, but only for insiders who have to file them.
• The 8-K or proxy covers executive grants. Alphabet's executive RSUs, for example, vest quarterly in 12 installments.
• Secondary sources (crowdsourced compensation sites, financial-planner guides) give splits and cadence for rank-and-file staff. They often disagree.

So an A usually takes two filings, and the cadence for staff below insider level usually stops at B.

Terms used throughout:

• Vest shape: the share of a grant that vests in each year, written as a split such as 25/25/25/25.
• Cadence: how often vesting events happen inside a year (monthly, quarterly, semi-annual).
• Cliff: a period before the first vest, during which a departure forfeits everything.
• Refresh: an additional grant made after hire, usually at a review cycle, on its own schedule.
• Stay-until rule: a clause that requires the employee to be employed on the payout date to receive a bonus.
• Clawback: the obligation to repay all or part of a sign-on bonus if the employee leaves within a set window.
• Post-termination exercise window: the time a departing employee at a private company has to buy their vested options.

Vest shape and cadence by employer

The new-hire grant's split and cadence decide how much a candidate forfeits at any given tenure, and employers vary a lot. On the reported splits, Amazon is back-loaded, Google is front-loaded, and Meta and the typical startup vest evenly.

The next table converts those splits into what has vested at the two tenures that matter most in sourcing. The 18-month column is derived and assumes even vesting within year 2.

Amazon

Secondary sources agree on the yearly split: 5% in year 1, 15% in year 2, then 40% in each of years 3 and 4. They disagree on cadence for years 3 and 4. One says quarterly on Feb 15, May 15, Aug 15, and Nov 15. Another says quarterly for L4 to L7 and semi-annual (May, November) for L8 and above. One source also describes "5% after the first year, 15% after the second and 20% every six months" as the schedule before a change, which is one more reason to hold the cadence at C. I did not retrieve an Amazon filing, so the cadence is unverified. Amazon pays sign-on bonuses in two installments across years one and two, which is the employer covering its own back-loading. That means the clawback date and the vest date have to be read together.

Google

Google moved employees globally from 25% a year to front-loaded vesting of 33%, 33%, 22%, and 12%, and one source dates the worldwide rollout to May 2021. A crowdsourced compensation site also lists other Google splits, such as 50/28/12/10. One secondary source says cadence depends on grant size: annual under 32 units, semi-annual for 32 to 63, quarterly for 64 to 159, monthly at 160 and above. Grade the split B, and only treat it as confirmed once the candidate's own vest table matches it.

Meta

The 10-K settles the length at four years. Insider filings put vest dates on February 15, May 15, August 15, and November 15, commonly one sixteenth per quarter. Meta's filings also show grants vesting one twentieth per quarter over five years. The quarterly date grid is reliable, but the split is not uniform across grants.

Startups

The typical pattern is 25% at the one-year mark, then 36 monthly vests. The cliff matters more than anything else here. A candidate at month 11 has nothing vested and everything to lose, while a candidate at month 13 has a quarter of the grant in hand and still faces the exercise window covered below.

How do refresh grants stack on top of the initial grant?

Treat every refresh as its own grant, with its own schedule starting on its own grant date, and add them up. If you count only the initial grant you understate forfeiture, and the error is largest at back-loaded employers.

What the public record documents:

• Amazon: each refresh follows its own 5/15/40/40 schedule, so by year 4 three grants can be vesting at once. A candidate at year 3 is in the heavy 40% years of the initial grant while also sitting in the light early years of later refreshes. Moving them is expensive on both counts.
• Meta: refreshers arrive annually in February and vest quarterly over four years. A secondary source notes the refresher does not factor in the value of existing unvested grants, so stacks grow faster than you would expect from the initial offer alone. Meta converts dollar-denominated grants to shares using the average closing price over the month before the start date, so the share count depends on that window and not on the price on the offer day.
• Google: refreshers typically vest evenly at 25% per year, according to a secondary source. That makes a Google stack a front-loaded initial grant with even-vesting refreshes on top.

The working assumption in the procedure below is one refresh per review cycle. Meta's February cycle is documented. For most other employers the review month is not public, so mark it unknown and ask the candidate rather than guessing.

Stacking only works if you start from a real tenure, and tenure is the input recruiters most often guess. Refolk returns people at a named employer by start-date band, so the stack starts from when someone actually joined and not from a title. For the back-loaded case, this is the search I would run first:

Bonus payout dates and stay-until rules

For each employer, record three things: the payout month, whether a stay-until-paid clause exists, and whether the bonus is discretionary or formula-based. Together they decide whether a candidate who resigns before payout loses the bonus, and so whether your offer has to replace it.

What is established:

• Timing. Compensation experts told WorkLife that bonuses are typically paid between January and March. Nobody has published the share of employers that pay in Q1, so record each company's month on its own and do not assume a quarter.
• California. An earned bonus is a wage under Labor Code 200. The only exception is a lawful, unambiguous "must be employed on payout date" policy where the employee left before that date. In practice, a California candidate who resigns before payout may forfeit the bonus if such a clause exists. Final wages are due within 72 hours of a resignation, unless the resignation comes with advance notice. The waiting-time penalty is one day of pay for each day of delay, up to 30 additional days.
• Discretionary vs formula. Whether the bonus is discretionary or formula-based affects whether it counts as a wage in California, so record it.
• Massachusetts. The state's highest court held that retention bonuses do not count as wages under the Massachusetts Wage Act.
• Other jurisdictions. Not established here. Check the governing state's wage statute before you rely on any rule.

What it implies for the window: the cheapest move is usually to start conversations before payout and set the start date after it. The candidate collects the bonus and your offer does not have to buy it out. A start date that falls before payout at an employer with a stay-until clause adds the full bonus to the buyout.

Does a post-bonus quit wave exist?

The evidence for one is thin. Forbes, analysing Federal Reserve quits data, found quits run higher in January and August. In a JobHire AI survey of 2,000 US workers, 59% expected bonuses and 48% of those said they planned to quit once paid. That is stated intent, not observed exits. Sources quoted by WorkLife said it is virtually impossible to put an exact number on bonus-timed resignations. Nothing published breaks departures down by function or measures the effect in weeks.

Sign-on clawback terms: window, shape, basis, triggers

Record four things for each clawback: the window, the shape (pro-rata, tiered, or cliff), whether repayment is gross or net, and which kinds of exit trigger it. Secondary sources put windows at 12 to 24 months, but the shape changes the cost far more than the window does.

Reading the filed examples

• Lumen. The employee owes the gross bonus less $12,500 for each full month worked after month 12, and nothing after month 24. A candidate leaving at month 18 owes the gross amount less $75,000.
• TheRealReal. A $1,000,000 sign-on, paid quarterly in four equal payments over the first twelve months, is fully repayable within 12 months and 50% repayable within 24. A candidate leaving at month 13 still owes $500,000. Tiers like this create a cliff on the anniversary date, so a one-week difference in start date can halve the buyout.
• Secondary norm. One source says 12 months is most common for bonuses up to about 15% of base, with larger bonuses carrying 18 to 24 months.

California AB 692

For California agreements executed on or after 1 January 2026, AB 692 (codified at Business and Professions Code section 16608 and Labor Code section 926) requires:

1. A stand-alone agreement.
2. A review period of at least five business days, and the right to consult counsel.
3. A repayment period of no more than two years.
4. Proration and no interest.
5. An option to defer receipt of the payment to the end of a fully served retention period.
6. Repayment only if the employee leaves voluntarily or is terminated for misconduct.

Because proration is required, the clawback cost for a qualifying California hire falls every month from day one. There is no 100%/50% cliff to time around. The law is not retroactive, and it is unclear how it affects remote workers outside California.

Private-company exercise windows

At a private company, the post-termination exercise window decides how much cash the candidate needs to keep their vested options. The standard is 90 days, but some companies extend it, and an extended window changes the buyout conversation.

• Why 90 days. ISO tax treatment under IRC Section 422 requires that ISOs be exercised within 90 days of termination. A VC quoted in Forbes said at least 90% of the companies he saw over a decade used a 90-day window.
• Extended windows. Pinterest moved to seven years for employees with at least two years' tenure. Quora, Asana, and Coinbase are also named among companies that extended.
• The tax catch. After 90 days an option no longer qualifies as an ISO. An extended window keeps the right to exercise but does not keep ISO treatment.
• Tender offers and secondaries. I found no public data on how these show up in public signals. Ask the candidate directly whether they have had liquidity.

For the buyout, exercise cash is strike price times vested options, plus whatever tax the exercise triggers. Have a tax professional estimate that part. On a 90-day window this cash is due almost immediately, and it is often what actually stops a startup candidate from moving.

How to build a pay-calendar row and a buyout estimate

Work through seven steps: classify the employer, pull the filings, grade against secondary sources, fill the bonus and sign-on rows, stack the candidate's grants, and produce a dated window with a buyout estimate. The order and the time estimates are my own proposal. No published source sets them.

Use this row skeleton for each source company on your sheet. Every field gets a value, a source, and a grade, or the word unknown.

Employer:
Public or private:
HQ jurisdiction:
Level band covered:

Vest length:            value | source | grade
Vest split (Y1-Y4):     value | source | grade
Vest cadence and dates: value | source | grade
Cliff:                  value | source | grade
Refresh month and schedule: value | source | grade

Bonus payout month:     value | source | grade
Stay-until clause:      yes / no / unknown | source | grade
Discretionary or formula: value | source | grade

Sign-on clawback window: value | source | grade
Clawback shape:         pro-rata / tiered / cliff | source | grade
Clawback basis:         gross / net | source | grade
Clawback triggers:      value | source | grade
AB 692 applies:         yes / no (signing date, governing state)

Exercise window (private only): value | tenure condition | grade

Row grade (lowest of the above):
Last checked:

Once the row and the stack are done, the outreach decision comes down to two numbers: how much unvested value the candidate forfeits on the planned start date, and how much clawback is still owed.

Before you price anything, get the candidate's own vest table. This is the message I send once a candidate is engaged:

Thanks for the conversation. To put together an offer that covers what you would leave behind, I need a few details from your own documents, not your employer's general policy:

1. The vest table for each grant you hold (grant date, number of shares, and the schedule of remaining vest dates).
2. Your bonus payout month, and whether your plan requires you to be employed on the payout date.
3. If you received a sign-on bonus: the repayment window, whether repayment is prorated or tiered, and whether it is calculated on the gross amount.
4. If your company is private: your option strike price, vested option count, and how long you would have to exercise after leaving.

Screenshots of the vest schedule from your stock plan portal are fine. Please leave out anything your agreements say is confidential.

Where pay-calendar values go wrong

Most errors come from one of eight places. In each, a company-level value gets applied to a person whose terms differ, or a source gets credited with more than it actually says. Each item below names the false positive and the check that catches it.

1. Vest split taken as company-wide. Tables of employer splits leave out non-standard grants. Google has several schedules on record, including 50/28/12/10, and Meta has 1/20 and hybrid schedules. The false positive is a 33/33/22/12 buyout for a candidate actually on 50/28/12/10. Check: the candidate's vest table.
2. Cadence assumed from one source. Amazon cadence differs by level and by source. A May/November assumption can put the window up to 3 months off for L4 to L7, where quarterly vesting is reported. Check: hold it at C unless a Form 4 confirms.
3. 10-K treated as settling cadence. The 10-K gives "four years", not the split or the dates. Citing it as proof of cadence is a false A. Check: grade length and dates on separate sources.
4. Refresh grants ignored. Counting only the initial grant understates forfeiture. Amazon can have three grants vesting by year 4, and Meta's refreshers do not account for existing unvested value, so stacks grow faster than expected. Check: one refresh per review cycle, confirmed against the vest table.
5. Bonus assumed safe after the performance year ends. Where a clear stay-until-paid clause exists, a California candidate who resigns before payout may forfeit the bonus. Check: the payout month and the clause, and the governing state.
6. AB 692 applied where it does not reach. Applying it to a pre-2026 contract or a non-California employee is wrong. It is not retroactive, and its reach to remote workers outside California is unclear. Check: the signing date and the governing state.
7. Clawback priced net when the contract says gross. Lumen's letter demands the gross amount. Pricing on net understates the cost by the withholding. Check: the basis line in the sign-on letter.
8. A 90-day exercise window assumed. At a company with an extended window, this overprices the exercise cash the candidate needs. Extended windows can also depend on tenure (Pinterest required two years). Check: the option agreement and the candidate's tenure.

The employer's policy tells you where to look; only the candidate's vest table tells you what to pay.

Underneath all eight is mid-cycle policy change. Google's move to front-loading and the Amazon schedule described as pre-change both mean that a grant's date can decide which schedule it is on. Record the grant date next to every split.

Who can confirm the mechanics, and keeping rows current

The people who understand how an employer's equity and bonus plans actually run are stock plan administrators and Total Rewards leaders, and they are concentrated at US issuers and plan administrators. Use them to review your own buyout model, not to extract another company's confidential terms.

In Refolk's index, "Stock Plan Administrator" accounts for 21 of 25 sampled US profiles, and the top US employers are UBS (3), Apple (2), and Ionis (2). UBS appearing suggests some specialists work at plan administrators rather than at issuers. The UK list leans toward retail (River Island with 2, plus Tesco, Bunzl, and Headlam), and I have not verified why. There are about 2.4 US Total Rewards leaders for every US stock plan specialist. If your own team has no one who can sanity-check a stacked buyout, Refolk can list Total Rewards heads or equity plan managers at plan providers by name. That is a hire or an advisor, not a source of a competitor's data.

Rows go stale in predictable ways. Re-check a row when:

• The employer files a new 10-K (service period, any disclosed grant changes).
• New Form 4 filings show vest dates that differ from your grid.
• A candidate's vest table contradicts the row. Their document wins, and the row drops a grade.
• The candidate's contract was signed under a different law than the one you assumed.