Google's Stock Couldn't Retain People, So Facebook Changed How It Grants Them?

Deep thoughts on AI and aspirations —— ByteDance Deep Thinking Circle

When Google went public in 2004, it created a batch of young millionaires. Side effects quickly emerged: once the options granted at hiring fully vested, some engineers cashed out and left for Facebook, which was still inconspicuous at the time but desperately recruiting talent. Silicon Valley calls this “graduating”—graduating from Google.

Facebook benefited from this wave of talent migration while simultaneously realizing something: when they went public, the same thing would happen to them even more intensely. Employees holding years of accumulated, nearly zero-cost restricted stock units would, once liquid, potentially trigger the largest arbitrage and talent migration in internet history.

Looking at these two events together reveals the real problem with equity incentives. It was never about “how much money to distribute,” but rather “how to split vesting across time.” Most companies focus on total amounts and percentages when discussing incentives; a few design the time structure. These two companies’ approaches happen to demonstrate both sides of this craft.

What Retains People Isn’t the Stock, It’s the Part They Haven’t Received Yet

First, let’s break down a basic concept. Granting and vesting are two different things: granting is a promise, vesting is fulfillment. The time gap between promise and fulfillment is where incentives truly work. Stock is just the vehicle; the time gap is the product.

This is where Google’s lesson lies. Silicon Valley convention allowed departing employees to keep vested options, so once that large initial grant fully vested, the retention rope was severed. With exercise prices near zero and stock prices up dozens of times, people left with paper wealth while the company retained nothing.

Facebook’s later design for executives went in the opposite direction. Incentive shares were granted annually, but already-granted portions had to mostly vest before newly granted shares began vesting, then vested over four more years. In other words, at any point, the “promised but not received” portion in someone’s hands was always greater than the “already received” portion. For ordinary employees’ RSUs it was even stricter: 25% of the total vested after holding for one year and six months post-IPO, with the remainder vesting monthly at 1/48. For the earliest grants, it took four and a half years from IPO to fully vest. The earlier the grant, the longer the lockup. This wasn’t generosity—it was precisely calculated punitive retention.

Here’s a risk most people don’t consider. The day golden handcuffs expire is a cliff. Once the four-and-a-half-year vesting period ends, the employee’s account is settled and the company has no new rope. Facebook’s design didn’t eliminate this risk, only postponed and smoothed it. The real solution is to have the next set of handcuffs on before the previous set expires: the standard for judging whether a vesting arrangement works is just one thing—on the day the last tranche vests, does the employee still have unvested future grants?

The Tool Is a Signal: Options Bet on Growth, RSUs Deliver Compensation

What tool you use says more about the company’s self-assessment than how much you grant.

The logic of options is leverage. Every dollar of appreciation above the exercise price is shared by employees, essentially making everyone bet on company growth. Using them during growth phases maximizes incentive efficiency. Google’s pre-IPO incentives represented about 28.5% of total equity, averaging 7.1% annually, with over 99% as options. Exercise prices before 2002 were as low as a few cents to two dollars, and the last batch in 2004 was set at 70% of the IPO price.

But options have a problem that worsens with valuation. After Facebook accepted Microsoft’s investment in 2007, its valuation reached $15 billion. With exercise prices set high, the upside for employees was compressed, and incentive efficiency plummeted. A high-valuation company granting options is essentially issuing lottery tickets unlikely to pay off.

Google’s trajectory is a complete sample of this curve. Starting in 2007, it used a two-to-one combination of options and RSUs, balancing leverage and stability. After the 2008 financial crisis deepened, the ratio shifted to nearly one-to-two, tilting toward stability. After 2014, it stopped granting options entirely, leaving only RSUs. Post-IPO annual grants stabilized at 1.5% to 2% of total equity, with incentive costs around 5% to 6% of annual revenue. The retirement of one tool marks a company’s transition from betting on growth to paying wages.

One detail deserves separate mention. Facebook’s large-scale switch to RSUs in October 2007 was primarily defensive: employees were exercising early for tax purposes and selling shares on secondary markets, threatening to breach the 500-shareholder regulatory threshold before going public. Switching to RSUs was forced by regulations. In hindsight, this forced move happened to match the needs of its high-valuation phase. My assessment is that correct tool selection is often forced by constraints; the difference is some companies learn to continuously reassess after being forced once, while others remain stuck after being forced ten times.

A word of fairness for employees. Many understand RSUs as “more stable,” but that intuition doesn’t hold for employees. Under U.S. Tax law, RSUs trigger tax liability the moment they vest, potentially requiring payment of taxes on depreciated stock during downturns. RSUs stabilize accounting and dilution management on the company side. The first-principles question in choosing tools is actually just one: do you want employees to bet on company growth or receive wages denominated in stock? The former needs leverage, the latter needs certainty. Mixing the two creates something that resembles neither.

When Markets Crash and Options Become Worthless, That’s When You See True Character

From 2008 to 2009, Google’s stock price fell 45%. Options granted post-IPO had exercise prices above market price, becoming “underwater” options. For employees, incentive value dropped to zero; for the company, previously recognized incentive costs were essentially wasted. This is the situation every option-granting company faces in bear markets.

Google’s 2009 approach is worth copying, but what’s more worth copying is its sense of boundaries. One-time exchange: employees could swap high-exercise-price old options one-for-one for new options near current stock price. Three accompanying conditions applied simultaneously: only price swapped, not quantity increased—old options canceled, new options granted; new options had reset waiting periods and extended vesting; the two founders, CEO, and non-executive directors were excluded from the plan.

The three conditions each address one thing. No quantity increase protects shareholders from second-round dilution, since investors were also suffering losses. Extended vesting prevents the exchange from becoming an arbitrage window. Excluding management preserves the fairness of the entire action. Most companies also do option repricing in bear markets, but the common approach has the company unilaterally bear additional costs while employees benefit at no cost—after doing this once, everyone expects bailouts during the next price fluctuation. The essence of Google’s plan is exchange: the company helps you cut losses, you pay with longer lockup.

Boundaries must also be clear. This design has two implicit premises: company value trends upward long-term, and a public exit path exists. Both premises require discounting in domestic markets. IPO timing is uncontrollable, private company old shares lack liquidity, and what employees “see” may take many years to become what they “get.” Real-world responses include cash-based long-term incentives, as well as virtual dividends or discounted buybacks of old shares for long-term holding employees. My view is that buybacks play the role of pressure relief valve in this time machine. Without a pressure relief valve, “seeing more than getting” gradually transforms from incentive to arrears. From day one, companies granting stock should think through where buyback money will come from.

One additional number: over 85% of pre-IPO companies choose to recover vested shares when employees leave, with buyback prices that can reflect returns but must be significantly below levels for current employees. Exit price design determines the quality of retention structure at the final moment.

The Pool Burns Faster Than You Think

Finally, let’s discuss resources themselves. Initial equity pools typically represent 10% to 20% of total company equity, with actual pre-IPO grants mostly falling between 10% and 15%. In terms of usage pace, pre-IPO grants average about 2% to 3% annually, with 60% to 70% of the total granted in the three years before IPO. In other words, the bulk of incentive resources is spent in earlier stages.

These numbers point to an easily misunderstood fact. Total planning is a static problem; pace is the dynamic problem. Insufficient pool size is a rare issue; more common is burning too fast early: using half the pool in Series A, then having to compress percentages or have all shareholders dilute again when hiring executives in Series B. Later grants cost less—with each valuation increase, fewer shares are needed to retain the same person. But later grants risk losing people. Granting early and granting aggressively—you can usually only pick one, the other must yield. This tradeoff has no standard answer, only one hard constraint: when the pool runs dry, are the people the company most needs to retain still within their vesting period?

Back to you, the reader. If you’re receiving an offer, don’t just look at grant numbers—ask “when does this incentive’s last tranche vest, and what happens if I leave in between?” The vesting schedule matters far more than the total. If you’re designing incentives, treat equity incentives as time structures, not total compensation to distribute. The test is one sentence: good design makes people feel they’re losing out by leaving at any point; bad design only makes people feel the company settled up on vesting day.

Methodological boundaries of this article: Only two cases—Google and Facebook—both survivors that weathered cycles and became giants. Their parameters can’t be directly copied. What can be copied is the structure: the mismatch between granting and vesting, the match between tools and stages, crisis remediation with boundaries, plus a pressure relief valve. Parameters must be recalculated using your own lifecycle.