Part 2: Compensation for Startups: Designing Your Compensation Philosophy
This article is part of a series on compensation for startups. It covers designing your compensation philosophy. The four parts are:
If you are actively designing and implementing any of these systems, we also have some Do-It-Yourself packages you can buy that include a recording of us explaining all of this in more depth as well as dozens of templates. You’re also welcome to reach out to hello@andthen.com with questions.
When a startup crosses 50 employees, founders start getting more questions like “I want to know what my career path is” or “Can I have a raise?” or “Why is that person making more than me when we’re doing the same job?”
For the founder, it can be overwhelming and frustrating. Overwhelming because you realize you may not actually have an answer to those questions – your compensation probably wasn’t super thoughtfully or methodically designed, and it may have taken on a life of its own. I see that a lot in early-stage startups.
And then it's frustrating because you might not even have a real business yet. You probably barely have product/market fit. So part of you is thinking, “Why are people asking me for raises when we might not exist in two years?”
It can both feel like a distraction AND a realization that there’s an actual problem you need to fix because:
The decisions you make around compensation impact people’s lives in a very real way.
You will perpetuate systemic bias if you’re not methodical about comp design.
Comp design can make or break your financials. The two most expensive things for software startups are usually servers and people — even the smallest comp decisions are about spending lots and lots of money.
This post shares the most important decisions you need to make to design a compensation philosophy for your startup – a philosophy is made up of the foundational principles that let you implement a methodical, scalable compensation system. I will cover the implementation part in a post that is coming soon. You can decide to get methodical about compensation at any size but I’ve done it a lot at companies that just crossed 50+ employees. Generally, the earlier the better because then you have less mess to clean up.
In addition to helping you understand the philosophical decisions you need to make, I’m also going to share the way I do this when I’m either operating or consulting. A lot of the system that I’m going to share is relatively common across companies in the tech industry, and I’m a big believer that your focus as a startup should not be on innovating in the HR world unless you have a very good reason. Within this system, you can make philosophical decisions that emphasize different things – Colleen McCreary’s work on role-based pay is one example. The goal of sharing this is to help you build something that works and then go back to the essential work of building a product people want and a company that will last.
Almost all the components of this system can scale for a while. It is loosely based on what we designed at Facebook around 500 employees and that system scaled to thousands of employees.
While I believe comp design deserves attention, I also just want to say that I’ve learned over time, again and again, that compensation isn’t what makes people happy to work for your company every day. Comp is honestly more likely to make people unhappy than happy, so truly, the best you can design for is basically to have people not think about it all the time. That is what fair and systematic compensation practices do. If something feels unfair, it grates on people every day, which just shortens their lifespan at a company. And at the same time, there are diminishing returns for tinkering with compensation systems. What you really want is a system that’s fair enough and clear enough that most people just put their feelings about compensation in a drawer and only worry about it once or twice a year.
The compensation system I describe below works in tandem with a leveling system and a performance review system (coming soon in a new post). Ideally, they are implemented together, meaning you decide on levels, design your compensation system based on those levels, and then use a performance review cycle to roll it all out and update it 1-2x per year.
The core principles of design
Fundamentally, I believe compensation systems should fairly reward people for their contributions to the company. The ideal system, in my mind, removes as much potential for bias as possible, reflects the company's stage and values, and emphasizes paying more for higher performance. The compensation system that I’ve implemented dozens of times is designed based on 3 principles:
(a) Market-based
Most companies today use a market-based salary structure, which means that their cash and equity compensation decisions are based on being competitive with other companies of a similar size and stage. Simply put, in a market-based comp model, you gather pay data from those companies and then decide how to position yourself relative to them.
Market-based pay is based on the idea that there are supply and demand curves for every job at every stage company. For example, engineers are generally in higher demand and lower supply than finance analysts, so salaries/equity grants for engineers are typically higher. There are infinite examples here including specific markets for specific types of engineering roles / skills, but the basic principle remains the same.
In a market-based philosophy, you look at market data on a certain cadence, like once a year, to confirm if your salaries are still positioned where you want them to be relative to the market. If the market moves, then you adjust your salaries to meet the market.
Until 2022, the market was moving up every year pretty dramatically for most jobs in tech, and most companies were offering 3%+ raises every year to meet the market. This year, I have heard a lot of companies choose not to adjust salaries or give raises, or only give raises to their highest performers. That is because the market has stopped moving up and might even, for some roles, be moving down. For context, this is the first time that has happened in my time (16+ years) in the tech industry.
Just for anyone who wants to go deep: I know this dynamic has been puzzling for some folks given the rate of inflation over the last two years – how can salaries not be going up at the rate that inflation goes up? Practically, this is the effect of multiple market dynamics happening at the same time. The hiring craze in 2021 led to big salary increases for a lot of roles, then hiring froze in the most dramatic way that the tech industry has experienced since the bubble burst in 2000. Demand froze and with all the layoffs, supply has dramatically increased. At the same time, inflation has been causing prices and the cost of living to grow. The very basic reason you are seeing static salaries in a lot of companies is that the first two dynamics – overinflated salaries and then a dramatic increase in supply coupled with a decrease in demand – are counteracting the inflationary dynamic. Ok, back to our main point…
In rare cases of the market swinging way up, like in 2021, you may need to give everyone an emergency midyear adjustment (upward) to stay competitive. I really don't recommend this, but 2021 was crazy! More commonly, this is needed for a couple of high-demand roles – AI/ML engineers and data science are recent examples of roles that have had rapidly moving markets. When the market for a role is changing rapidly, sometimes you need to do something short-term in order to retain or attract talent. In general, though, I strongly advise only looking at / adjusting your compensation data relative to the market once a year, but more on that below.
Final point: with a market-based approach, you don’t typically worry about things like cost-of-living increases, because you plot your salary bands or points and then have a geographic multiplier that takes the cost of living into account. You might do periodic increases to match the market rate, but not cost-of-living increases.
(b) Make everything as formulaic as possible
I strongly believe that if you do nothing else, you need a compensation system that is fair. If you don’t have a system that’s grounded in a set of principles, then you’re just making comp decisions based on what you had for breakfast or what your managers had for breakfast and that typically leads to A LOT of bias. And if you don’t believe that most compensation systems are biased, please go see the data on pay for women. Anytime I look at an early-stage startup’s compensation data (before they implement something methodical) I see bias.
To achieve a more fair system, I always recommend being as formulaic as possible. What does that mean tactically? The more that you can have a spreadsheet with formulas spit out numbers you don’t modify, the less likely you are to introduce bias into the overall system. Design your comp system to eliminate as many points of manager discretion as you can – manager discretion almost always = bias. Manager discretion can look like everything from giving managers a pool of money and letting them decide how to distribute that pool amongst their employees for raises or bonuses all the way to something like allowing managers to modify what the formulas spit out. You can have discretion in your system if it feels important but just know that the more discretion you allow, the more likely you are to have bias.
I also want to point out that people always find out what other people are making. This is one of the things that newer managers get surprised by but it is always true. If there is bias in your system, everyone will find out about it. There’s nothing more draining as a high performer than knowing that the person next to you, who’s not doing as well or as much as you are, is making 1.5x what you do.
Formulas prevent issues all along the employee lifecycle, from recruiting to retention. One of the big mistakes I see startups make is to over-recruit: if you don’t have a principled compensation system, and you’ve spent a lot of time recruiting a candidate, you might stretch your compensation framework just to get them to join. You end up with an engineer making 1.5x what others around them are making just because they had a competing offer from [insert cool company here]. What happens when this person isn’t performing as well as others? Are they really 1.5x as valuable?
If you have a principled compensation system, it allows you to have comp conversations early in recruiting and to say “no” when candidates demand crazy salaries. It ensures that people are joining for the right reasons AND that the whole system is fair to everyone, not just the folks who had offers from Google or the people who are good negotiators. There are many many many crazy stories from the Facebook / Google wars of engineers going through the recruiting process at Facebook ONLY to get a counteroffer from Google. Think of all the wasted time and energy for everyone on all sides because of those compensation decisions. Don’t put yourself in that position.
Being formulaic also allows you to be transparent. Comp transparency could be an entire blog post unto itself so I will spare us all and just say: I don’t advocate for full compensation transparency, but being formulaic gives you the choice to decide what to share with all employees. If you have a formulaic design, then ideally all your conversations and battles with employees will be about defending the design of the system, not the specific decisions related to individuals. I do advocate for being transparent about the system.
(c) Pay for performance
I believe companies should pay high performers more. I was raised this way at Google & Facebook, and I generally think it’s a mistake for any company to undervalue its high performers.
I like to use a performance system to formulaically inform raises, equity increases, etc. In order to make “pay for performance” work, I basically tie a startup’s performance review system to its compensation system. Most performance review systems that connect to compensation systems have two outputs for every employee:
How did the employee perform over the last 6 months?
Should the employee be promoted to the next level?
The output of #1 is a rating, and the output of #2 is a yes or no.
For #1, the ratings are typically words like “Exceeds expectations” or “Meets all expectations” and those words typically correspond to a number. Without going into too much detail (will cover it in the perf review article), a lot of startups either implement a 5-point scale or a 7-point scale.
The way that you design compensation adjustments to connect to those outputs from the performance review cycle is a set of multipliers and rules. Behind these numbers are a set of decisions and values around how you want to reward high performers and the messages you want to send to folks. For example:
Is average performance good? If most people are getting a “Meets All”, then you probably want to make sure that getting that rating “feels good” from a compensation perspective by providing a solid raise for that rating (in the example it’s 3%).
How much do you want to reward high performance? Facebook had a very steep curve here for a very long time so if you got higher performance ratings you really benefited from a salary and equity perspective — in other words, multipliers got higher faster for the higher performance ratings. (In the example it’s 7-15% salary increase, plus equity for the highest ratings).
How do you want to handle folks that are performing below average? Small % raise or 0% raise? Managers may be disincentivized to give this rating if it means no raise, but when you’re more cash-strapped, you may prioritize your highest performers.
How do you want to think about additional equity grants for performance?
There are lots of other decisions to think about and make, and I talk more about how to execute this system in the perf post, but those are some of the big ones.
Ongoing equity grants are another big topic – including new hire grants, performance-based grants, and equity refresher grants (which are based on tenure). There are a lot of good articles on the internet about designing equity programs, so I won’t try. For example, check out Henry Ward’s writeup of Carta’s equity refresh program.
A quick side note on my feelings about bonuses
Since we’re talking about pay for performance, I will state my feelings about bonuses. Except in the case of sales compensation, I really strongly suggest not implementing any form of bonus structure in a startup. Why?
Bonus systems typically have two components that dictate how bonuses are calculated: individual performance and company performance. And startups aren’t good at measuring either one systematically. For startups, I recommend first implementing a performance review system for about 2-3 years – enough to be confident that you’re good, systematic, and fair about measuring individual performance. Then you can consider implementing a bonus. I actually implemented the bonus system at Facebook, but that was at around 750 employees.
The other issue is, as soon as you put a bonus in an offer letter or as soon as you give the first one out, basically, everybody assumes you're going to pay them a bonus every year at a hundred percent or more. They plan around it. You think a bonus is optional; employees think of it as part of their compensation. If you want a fast way to piss people off, then give them less than they expected for their bonus (or none).
Suffice it to say, bonuses are complicated to implement and execute. I would highly recommend just focusing on getting consistent and fair about performance reviews, and salary and equity increases before you tackle bonuses.
In case you feel you need to have bonuses, I would say simple is best. Every bonus system has downsides so unless you’re very confident about the behavior you want to incentivize, be careful. In case you want to implement a company-wide bonus, here’s the formula that Facebook and Google used back in the early iterations (again implemented at close to 1,000 employees and long after their businesses were well established):
Target % of salary (based on level) x individual performance multiplier x company performance multiplier x % of year worked This formula means the bonus is based on individual performance (as defined by a multiplier based on the rating they received) and company performance which is typically a multiplier decided by the CEO and/or the board.
Design decisions for your compensation philosophy
Assuming you are designing your philosophy based on the principles above, then the first thing to do as you create a compensation framework is to make a series of decisions that reflect your company’s philosophy on compensation. Those decisions will inform the design of your system. Here’s a list, and below it I’ll show you an example of how a hypothetical software startup might answer these questions.
Leveling
Note: You need a leveling system in place for this comp system to work. See the leveling post for more in-depth design questions and templates
How many levels will you have?
Will management and IC work be paid the same? (Strongly recommend having ICs and Managers be paid the same. You do not want to incentivize people to become managers.)
What titles go with each level?
Position relative to Market:
Where do you want to position your company’s comp offerings relative to other companies your size? Top, middle or lower half of market?
If you are below 75th percentile, it is worth asking how will you handle recruiting from / competing with the FAANG companies.
Where do you want to position your cash vs. equity offerings?
What percentile for cash? What percentile for equity?
Will you negotiate salary with new hires? What about existing employees who come to you with offers from other companies?
Location-based adjustments
How will you handle pay in different geographies?
In different cities/states in the US?
What about in different countries?
What happens when someone moves (either to a more or less expensive market)?
Salary/equity bands
Will you offer salary/equity ranges (bands) and negotiate within them, or a non-negotiable amount for each discrete level+role (points)?
Will your bands overlap? (Can a high-performing IC3 get paid more than an average-performing IC4?)
Pay for performance
Will you give bigger compensation increases (salary & equity) to high performers?
Do lower performers (below average) get raises? Equity grants?
What is your process for making exceptions to any part of this system? Who has to approve them? (not communicated broadly but something you have to decide)
Scheduling
How many times per year will you adjust employee compensation?
For general salary raises & equity refresher grants?
How many times a year will you do promotions? Will you adjust both equity and salary when you promote people? (Some companies promote twice a year but wait to adjust equity and just do it once a year.)
In which months does this happen?
Example: A hypothetical startup’s comp philosophy
The bullets above are key questions you need to answer to create your own compensation philosophy. Typically that philosophy is captured in a document that you share with employees on your wiki. It helps ground all compensation discussions and managers can refer people to it. My team created a template version for you based on what we’ve done for previous companies.
This startup’s basic decisions based on the questions above are as follows:
We use the exact leveling system in this leveling article
It has 8 levels with ICs and managers as parallel tracks, meaning compensation will be the same regardless of whether you’re an IC or a manager. We do not want to incentivize people to become managers.
Salary and Equity are both positioned at the 50th percentile of market for all roles
Most startups should either use 50th or 75th because they are conserving cash (and cash shouldn’t be why employees join startups; quite the opposite, employees should join startups because they believe in the upside of the equity). You can choose to position higher relative to the market, it’s just expensive.
Part of positioning at 50th means that we don’t expect to compete with FAANG employees on cash. If someone is in our recruiting process either from a FAANG company or is also in process with a FAANG company, we are going to have several very explicit conversations about compensation and specifically salary since we can’t compete.
We only have US employees, and because we are a fully remote company, we pay all employees 100% of market (at the 50th percentile) regardless of location.
We’re paying people in San Francisco and Kansas the same amount, rather than adjusting the Kansas salary down from the 100% point. This is the simplest way to handle comp as a remote company right now, but that doesn’t mean it’s the best. It means we don’t even have to answer the question about what happens when someone moves. We’ve also simplified our lives by only hiring in the US.
Geographic compensation is one of the most complicated parts of the comp world right now because it is moving and changing quickly. Just like comp transparency, this could be a whole article.
We pay for performance
This startup will give larger year-end salary and equity raises to people who get high ratings during perf review cycles. See an example of how to set this up in the perf post.
We have salary/equity bands instead of points to allow for more simplicity and because we don’t believe our recruiters and managers can handle a “no negotiation” approach.
You can use points instead — points assign an exact salary amount for each level, no negotiation – but it takes a lot of discipline to stick to them so most startups use bands because they’re easier to enforce and still lead to more fairness than no bands at all.
Our bands overlap to incentivize high performance
The main benefit to overlapping your salary bands is that you are not incentivizing people to get promoted in order to be paid more. If your bands overlap, sometimes it is better for an employee to be higher performing at a lower level than to try to increase the scope of their role.
We review everything about everyone’s comp 1x / year in January and do promotions 2x / year (Jan and July)
These are tied to June/July and Dec/Jan performance review cycles.
Once we’ve made all those decisions, we are ready to document our compensation philosophy and start implementing our compensation system.
We’ve created a template compensation philosophy document to help get you started with yours.
Once you have done the hard work of deciding on your compensation philosophy, the next question is how you turn it into reality. I am going to cover implementation in a separate post coming shortly as well as a final post on the performance review system that I typically design and implement. Stay tuned!
But primarily, as I said above, the purpose of this post is to help you see what it can look like to design a compensation system around fairness and consistency. If you have a system that you believe is designed to be fair, you can be clear and direct with employees about how it works, which can let them go back to focusing on building a company that will make the world better in some way. And, hopefully, all that hard work will turn into valuable equity for everyone.
This article is part of a series on compensation for startups. It covers designing your compensation philosophy. The four parts are:
If you are actively designing and implementing any of these systems, we also have some Do-It-Yourself packages you can buy that include a recording of us explaining all of this in more depth as well as dozens of templates. You’re also welcome to reach out to hello@andthen.com with questions.