Compensation Primitives
Or how to ensure employees are aligned with corporate strategy.
Introduction
When building something new, it’s often surprising just how many assumptions are inconspicuously baked into the various systems with which we interact every day. In this specific case I’m referring to compensation structures, an area that frankly doesn’t get nearly enough ink as it should. Nearly every startup in Silicon Valley offers a similar package of high salary and competitive equity ownership of the company, so what is there to write about?
This knee-jerk jump to the default belies the fact that this structure was not always the default; it emerged from a series of decisions, built on now neglected assumptions, that may not necessarily be optimal for the construction of your company. Take the salary – generally viewed as “superior” than its alternative (the hourly wage), but this conflates the fact that the employers offering fixed salaries will generally pay higher wages by nature of the company’s success, not because the salary itself is (necessarily) the superior compensation vehicle.
More senior individual contributors often find that they’ve unlocked a new wage ceiling when they shift into a contractor role, which shouldn’t be possible if the salary is de facto superior to the hourly wage. What’s missing with a simple “salary > hourly wage” heuristic is the context - under what conditions is this logic statement correct? Is it in fact universal? Specifically here, we can quickly arrive at the conditions that would yield a higher wage ceiling for the hourly wage — companies orient their wages around a 40ish hour work week but regularly extract 50+ hour work weeks, thus saving the company money but capping the given worker’s earning potential.
The fixed salary, then, is in most cases a tradeoff between a guaranteed wage floor and a fixed impact of actual time spent working. There is no 1.5x overtime pay in a salaried job, even though there are effectively no companies within the Innovation Economy that actually operate with a 40-hour work week, and thus most tech workers provide a level of wage subsidization to their companies in exchange for the certainty of a guaranteed wage.
This longer-than-expected digression provides some context for why defaults can be quite insidious when building novel things — without rooting out the foundational assumptions, it’s all too easy to create incentives that almost certainly equilibrate in misalignment. So how should we move beyond the defaults when considering the optimal compensation scheme for a new venture?
Compensation Primitives
Bezos-inspired views are almost certainly over-extended into inappropriate contexts, but one particular focus of his that I find particularly useful is the hunt for primitives. These are effectively the basic lego blocks undergirding a given service/platform that are, conceptually at least, modular and recombinable to meet the needs of many different customers. AWS was constructed (at least apocryphally) on a foundation of primitives, breaking down core IT needs for compute, storage, and others into their fundamental elements.
Worker compensation certainly benefits from a primitives view, though as I asserted from the jump, we too often conflate distinct primitives as a singular default, or we misapply a primitive given a distinct set of incentives we wish to achieve. Let’s first break them down here before how they can be well- or mis-applied
SALARIES provide a fixed, guaranteed floor for one’s income, typically quoted in yearly increments. Because of the extended time horizon, earnings here are largely decoupled from actual hours worked. The oft unspoken acknowledgement is that the job entails 40-50 hours per week, typically communicated through the more passive “our working hours are…”, but because hours aren’t typically tracked, there are rarely formal penalties for violating temporal floors, and certainly aren’t (direct) rewards for working more1.
HOURLY WAGES also provide a fixed, guaranteed payment floor, but unlike salaries, these rates are tightly coupled to time. As such, one’s earnings are linearly related to time spent, and there is no compensation without “putting the time in”. The flip side to this occurs when work within a given week extends beyond the agreed-upon “normal limit”, after which earnings are accelerated and hustle is rewarded 1.5-2x (yes, I’m referring to Overtime here).
You’ll notice that both of these compensation primitives are tethered to our inputs – the wages are guaranteed as long as we put in the work (effort + time) – and that performance does not matter2. Given this characterization, we then have two further primitives tied to worker outputs and thus to performance.
BONUSES are effectively “employee dividends”, tied in most cases to some basic performance-oriented goals (both individual and company), though in more extreme examples are tied to managerial discretion3. Sales commissions are the more formal and extreme version of this primitive, though they are occasionally manifested as the next form.
PROFIT SHARES remove the qualitative ambiguity of bonuses and instead tie compensation directly to an objective calculation of company profits. These are quite rare, though as I’ll discuss shortly could be a powerful addition to some company’s arsenal.
Where art thou, equity?
There is one final compensation primitive that doesn’t fit as neatly into this inputs/outputs structure, and it’s the one with which I launched the piece — that is EQUITY SHARES. In tech circles, equity shares are the norm; an absolute expectation such that any offer lacking in equity might be rejected outright. But tech is not a representative industry, and though I don’t have the exact stats on hand, I suspect that far less than 50% (and probably closer to 10%) of salaried non-tech workers hold equity shares in their employer.
Equity is different than our other primitives because it represents paper wealth – illiquid promises of future windfalls if both of two conditions are met:
The company appreciates in value significantly above the given employee’s strike price.
The company exits (either via IPO or acquisition)4.
So that’s it – the five base compensation primitives. This may feel a bit reductive – few reading this will find the primitives revelatory – but that is precisely the point. Primitives enable us to effectively reduce a problem set to first principles, lest we wish to doom ourselves to incorrect assumptions and misapplications.
Applying our Primitives
Let’s briefly look at three different examples to unpack why the given primitives were chosen for the given context.
Sales Commissions: Not One Size Fits All
Every B2B company offers some form of Salary + Commission, and the actual structure of these commissions fundamentally governs both what the company values and how it wishes to operate (whether this is intentional is another story…). In 2012, I joined AdColony, an early purveyor of mobile video advertising, where my colleagues and I had the good fortune of riding the largest growth wave in tech history (mobile) and arguably its largest benefactor (mobile gaming).
The macro-economic pace was so fast that we literally achieved 50+ weeks of unbroken week-over-week growth, built on the backs of startup mobile game studios that one after the other would leap from unknown to massive success – and that success in mobile games meant massive advertising budgets for us.
Given this type of environment, how should a commission be structured? The default here is also the simplest, which is to tie each seller’s individual commission to the ad spend of all clients she specifically brings in. This could be correct incentive alignment under three conditions:
The distribution of account revenues is relatively linearly distributed.
The securing of new accounts is largely skill-based.
Account revenues are largely driven by the sales action.
In AdColony’s case, precisely zero of these assumptions was true:
Game studio ad spend distributions were Pareto or power law distributed.
Once our brand was established, and because of our SDK-based solution, we became a default solution for most/all advertisers and thus each successful company would inevitably work with us, regardless of the salesperson5.
Because of the variable nature of performance ad spending, account size was derived primarily from “network quality”6, not from sales quality.
Given the violation of the three base assumptions, deploying the individual seller-focused commission plan should be rejected. But what if it wasn’t? The Pareto distribution, coupled with the luck of “landing” a top account, would mean that a bulk of commissions flow to just one person who may not actually be good at the job. And because of this “lucky takes most” dynamic, team members would be far more likely to snipe and undermine top account owners in an attempt to to steal these accounts, rather than focusing on growing the whole revenue pie. Not a great cultural move, and certainly not great for long-term growth prospects for the company!
We did, thankfully, deploy a compensation scheme that acknowledged both the reality of ad spend patterns and the growth strategy of the company. Rather than anchoring comp to individual accounts alone, each geographic region was comped on its collective revenue performance. This extended one layer down from the sales team to the account management layer, given that most of the account growth occurred after the first buy and required the joint efforts of our sales and account management teams. No single person was responsible for success, and thus communal compensation was the correct call.
Typical Tech Equity
Working long enough (or perhaps solely) in the tech sector often leads one to believe that salary + company equity is the only comp structure available, and thus the overwhelming majority of startups start here. But one really should first assess why this two-primitive structure dominates, and only if their new company shares these characteristics should it be adopted.
Tech companies are, almost definitionally, entities created for immolating cash, especially early in life. With cash so precious (especially when interest rates are non-zero), companies often can’t pay market rate salaries in cash alone, and thus they require a form of compensation that costs nothing now, may cost something in the future, and is enticing enough to attract talent with clearly suboptimal immediate term earnings potential.
Equity shares thus offer the company a 0% loan derived from the delta between the employee’s accepted salary and her actual market rate, multiplied by the number of employees at the company. For the employee, equity shares represent a probabilistic opportunity for a significantly larger payday — sacrifice a bit upfront for a larger share of the back end (if the company actually succeeds7).
Cash + equity comp necessarily means that the company should be strategically geared toward an exit, and employees should be maximally focused on how to make this reality.
My quibbles about defaulting to an exit-first strategy aside, the salary plus equity primitive combo is pretty well-aligned with many growth companies. But must it be used for all? Should it in fact be the default, with nary a disagreement?
I’m obviously posing the question precisely because the answer is no, and the contrarian take comes in the form of a hyper-visible manifestation – OpenAI.
Massive and Profitable?
The creator of ChatGPT, founded by a former head of tech royalty (Y Combinator), has raised tens of billions to lead the AI “revolution”8. OpenAI represents the quintessential massive bet for which Silicon Valley is renowned, and yet the company offers no equity compensation. This is not to say its compensation package isn’t robust! A starting AI researcher receives a base salary of $300k, but the upside component is not equity but profit shares.
Why would this highly unprofitable enterprise build its entire comp structure off of profit shares? For one, OpenAI is itself a “capped profit” entity, and its parent company is a nonprofit, which certainly impacts equity ownership consideration.
More importantly, by not offering equity, the company signals that it never intends to exit, which fundamentally differs from the traditional startup and thus renders the value of traditional equity shares moot. Thus without equity in its arsenal, it requires a separate probabilistic offering in the form of profit participation units (PPUs). By basing its comp package on profits9, OpenAI directly communicates to current and future employees that its intention is to achieve profitability on a “reasonable” timeline for such shares to be realistically meaningful for the new prospective employee.
Conclusion
We spend quite a bit of time in this newsletter on the esoterica of corporations, and compensation is no different. But esoteric does NOT mean unimportant – especially if one wishes to create innovative new structures of innovation (as I do). Compensation structure must directly align with a company’s strategy, and mindless deployment of defaults is a recipe for premature company closure.
Compensation innovation is not without challenges! For one, it’s difficult enough to convince others to join a new venture, especially one that looks quite odd from the outside. Couple that with a comp plan that also bears little semblance of what one has seen before and…
Nevertheless, my job is not to cow to convention for convenience but to choose correctly for our present and future team members. If you have thoughts or examples of really creative compensation structures that ideally blend with the given company’s strategy and operations, please send me a note!
It's beyond the scope of this piece to explore the best ways to accelerate one's career advancement. implied of course is that working longer is, generally, a necessary strategy.
Of course one's ability to retain a given salary is performance dependent! But as long as one is actively working, that salary is rarely if ever docked.
See Michael Ovitz and early CAA.
Yes yes, secondary equity markets, blah blah blah. These markets provide small minority liquidity and are irrelevant for the vast majority of equity holders.
Though I will acknowledge that the speed of adoption was at least partially salesperson-dependent.
A combination of the quality of our algorithm and our supply.
Though this admittedly means that most employees lose value, given that just 11% of startups actually exit.
I'm with Ben Thompson in clarifying that AI is much more likely to be a "sustaining innovation" than a disruptive new platform.
If you read through the above link, you can see that OpenAI estimates the PPUs will be worth an additional $500k per year, or 62.5% of total compensation).