Defining roles

Great role definition mixes what the research has tested with what you only learn by doing it long enough to “just know.”

Great definitions

The most important thing to understand about a great role definition is that it is a rubric. Down the road, after definition, members of your team will collect evidence each candidate does or does not have the competencies you’ve laid out. You’ll do this through research, interviewing, take-home assignments, informal walks or meals, and reference interviews. Each person on your team must be able to mechanically and consistently score candidates based on your role definition document. Why? Structured judgment works: assign each person a set of competencies to evaluate and have them score them (1-5) independently, in writing with a sentence or two of evidence per score. But when it comes to combining the evidence and deciding, humans are hot garbage: add the scores up and let the total decide, because the moment the debrief meeting starts “synthesizing” holistically, you have reintroduced exactly the failure the research documents [2].

A great role definition is also formal: written down, the product of a deliberate procedure, not the hiring manager’s off-the-cuff sense of the role. When researchers measured which interviews actually predict job performance, the winners were structured interviews: built on a real, written definition of the job, with standardized questions and scoring [3]. Structured interviews predict performance about twice as well as interviews run off someone’s rough sense of the role [4]. And the structure starts with the written definition. When researchers taxonomized what makes an interview structured, a formal definition of the job was the foundation [3].

Great role definitions are outcomes-based, not function, activity, or generally desirable human traits-based. Start with asking yourself what someone in the role you’re defining will need to accomplish in twelve to eighteen months in the future. The novice version is “we need a head of sales to run the sales team.” The expert version is “build the sales team from two individual-contributor reps who report to the CEO into its own organization capable of driving $XM in revenue, complete with operations, predictable forecasting, standardized compensation and quota, one or two reps consistently exceeding quota, and a hiring pipeline so we can scale the team to 20 within eighteen months.” Same hire. Wildly different clarity. Specific beats vague here, and not as a matter of taste: specific, demanding goals are one of the most replicated results in organizational psychology [5], and the outcomes you write down become the goals your new hire gets managed to [5]. Three to five outcomes, not twelve. When you can’t forecast a number yet, write a milestone (“a repeatable outbound process exists”) instead of false precision. A rubric built on fictional numbers is worse than a vague one, because it gives you a false sense of rigor. The test for every outcome you write: could two people independently agree, at month twelve, whether it happened?

They are derived, not borrowed. A chef does not cook from a recipe found somewhere and neither should you. ChatGPT, Claude, Google, and job descriptions from other companies can be valuable tools when defining roles, but they cannot be your only inputs. In addition to your company’s specific outcomes, a great role definition takes into account the necessary Competency Stack based on the level (C-suite, VP, director, manager, IC, etc.) of the role, your company’s values, your stage of company (at least by revenue and headcount), and the specific nature of the team they’ll be joining. Use an off-the-shelf job description written by someone else and expect an off-the-shelf hire designed to fit someone else’s company, not yours.

A great role definition is properly scoped. You have to ask yourself if this role should even be a role. Is it one, two, full-time, part-time, or contract?

A role definition must also be complete. It must outline what the new hire will accomplish (outcomes), how they will be evaluated, and what they will have independent authority to decide. The best way to write the authority section is two columns: day-one authority and full-charge authority, with the explicit condition that moves each item from the first column to the second (“after your first successful quarterly close, you own the forecast”). A definition document that lists an authority the hire won’t actually hold on day one is misleading at best and a recipe for resentment at worst.

But role definitions also need to be honest about human capability. No one is perfect. All people have flaws. Outlier talent will come with weaknesses. Truly great role definitions include and differentiate between must-have strengths and acceptable weaknesses.

And finally, the best role definitions act to close feedback loops collaboratively and transparently instead of opaquely. While the job of a role definition is to give managers a rubric for evaluating whether they’ve found the right person, it’s also to create alignment between a manager and the new hire. Use role definition documents to close the loop at a few different moments. First, ask each new hire to write, in their own words, what outcomes they think they’re responsible for, then sit down together and compare notes with the original role definition document — adapting an exercise Peter Drucker prescribed for one big reason: people fail in new assignments mostly by continuing to do what worked in the old one [6]. Second, pull it up at day 30, 60, and 90 to make sure you’ve found the right person. And finally, periodically check to see if it needs updating or editing, and make sure to communicate about any changes up, down, and across.

Great definers

The most important quality in a great role definer is patience. Role definition is an investment, and like any investment, it is subject to the J-curve. People avoid investing time and resources into role definition because they can’t see the return on their investment for a while. Trust that the payoff will come later, when the right person is in the right seat, avoiding a mis-hire and costly replacement process. Patience is essential because great definers know it can take a good bit of time to know what great looks like, a process that takes calibration calls and (often) trying the job yourself.

Truly excellent hiring managers excel at making the squishy legible (this one is so critical, it has its own hiring competency). The things you most want (and don’t want) in teammates are difficult to name. It is okay to prefer working with people you like being around. The things you prefer, appreciate, and admire about other people — in addition to the things that drive you nuts — matter. Great hiring managers take note of these positive and negative emotional cues, examine them more closely, name the specific behaviors that are either productive or counterproductive, and teach their team how to filter those qualities in or out. “I don’t think they’re a culture fit for some reason I can’t quite put my finger on” turns into “You maintain calm poise in stressful situations” [7].

The best leaders know who to include. Role definition is never a solitary process, but it also should not be a consensus one. Every new role has stakeholders. Collect input deliberately and widely, but the final call stays yours. The marketing team has opinions on sales, engineering has opinions on design and vice versa. Some of their opinions will be good, and others less so. Founders need to find a way to be involved in hiring without suffocating the leaders they hire. And leaders need to include their bosses and their peers. No department is a silo without someone making it one. Don’t be that person.

Great definers also hold a high bar for the relationship. Every role definition implies a relationship, and the relationship is a two-way street. Before the search opens, the hiring manager needs to answer one big question from my own hiring checklist: are you enthusiastic and realistic in your commitment to develop a relationship based on openness and trust, where both manager and direct report seek ways to help and enhance each other? Would you be excited to get lunch with them if they were in town? I see this go wrong constantly — managers who define a role, run the process, make the hire, and never intend to invest in the person on the other side of it. Or the new hire, deep down, doesn’t respect and admire their boss, so they undermine them. The definition document describes a job. An employee signs up to work for a manager. Don’t skimp on either.

Greatness here also requires each hiring manager to know they work within a system. Leaders, to a point, are entitled to lead how they want to lead, but each leader operates within the boundaries their company has defined more broadly. A company’s values (the company-wide layer of the Competency Stack) predate most managers at a company, and it is the responsibility of each leader to filter for them to ensure talent compatibility.

A real nuance great definers hold well is defining the transfer of power. After all, you can only delegate something if you’ve defined it up front. We hire people to do things someone is often already doing, even if poorly. To define a role is also to define how and when power — decision-making power, payroll-button-pushing power — will be fully transferred to the new person, and the right answer is rarely all of it on day one. I recommend using Tobi Lutke’s Trust Battery [8], the idea that someone on day one is likely at or around 50% of a “battery” charged with trust. Only by “charging” the battery fully do all of the role’s responsibilities get released. A lot of folks bristle at this, because it inherently communicates a lack of trust. But some things are worth protecting until you’re sure you’ve found the right person. Make sure to communicate about the conditions for transfer: that’s what the two authority columns in the definition are for. It’s okay for you to privately hold a measure of trust, but written conditions are essential (“own the forecast after your first clean quarter”).

Great definers recognize that things change. Startups move fast. A role defined on June 1st may have changed by the time someone starts on August 1st. I always think of Alfred Korzybski’s quote, “A map is not the territory it represents” [9]. Role definition documents are “maps.” The job someone does becomes “the territory.” Done well, our maps capture the territory well, but all territory is a bit different. Great managers (and their teams) know that things change, they notice when they do, they talk about it, and then they update their maps.

And while these last two traits belong to the evaluation phase, great definers own them anyway, because they close the loop back into definition. People great at defining roles also have a willingness to cut losses. A Talent Partner at A16Z once told me that the best hiring process in the world will still fail 30% of the time, and for executives that number is closer to 50% (executives are better at interviewing, the theory goes). I’ve found it to be reliable napkin math, and it isn’t far off what Peter Drucker said: one-third right, one-third minimally effective, one-third outright failures. No other area of management, he wrote, had such a poor record [6]. Hiring the wrong person is costly. It is even more costly when you keep them around for a year, let alone multiple years. Great leaders know that the first ninety days are an assessment period to validate or invalidate whether someone is capable of excelling at the role crafted for them.

But of course, great hiring managers also update the system upon failure. To have hired is to have fired, laid off, and lost sleep over it. Unless you’re a psychopath. All great leaders, upon letting someone go, sit down and think about what went wrong in the hiring process. What part of the role was ill-defined? And they use that to improve the process the next time around. And do the post-mortem with the team that helped build the rubric. They wrote it with you, and they should see whether it predicted reality.

At Management Craft, we chart each competency on an abridged version of the Dreyfus Model, a system developed in the 1980s to chart how humans develop skills. The full model has six rungs (novice, advanced beginner, competent, proficient, expert, and master). For simplicity’s sake, each competency has a simplified rubric with only three levels: novice, competent, and expert. Use the rubric below to self-assess, use it as a team to calibrate each others’ capabilities, or use it as a conversational aid in the role definition process.

Founders are particularly prone to what I call “cowboy hiring.” They meet someone informally, return to the office as if they’ve fallen in love with the marketing equivalent of a French romance, and they declare their intent to bring them in to interview for a role nobody has ever heard of, let alone defined. Meeting someone great and then hiring them isn’t the sin. If you meet a brilliant marketer and you already have the marketing role defined, bring them in. The sin is having no definition at all. Professional exuberance is a rare workplace delight, but skipping past role definition is playing with fire. You might get lucky, but you might also burn the house down. Thankfully, you can prevent startup fires, and you can do so by writing a fucking role definition document.

1. Realize and accept you need to improve. Getting feedback is rarely easy. If you’re reading this, it’s probably because someone suggested that you have room to improve here.
2. Price the consequences of staying the same. Kurt Lewin sketched the foundational model for how people change back in the 1940s, and Edgar Schein later built it into the version managers still use [10]. To dust off those nasty old habits, our motivation to change must outweigh the comfort of staying the same. So put a dollar amount on each hire. A $150K hire you replace after twelve months didn’t cost you $150K. It cost you the salary, the recruiter, the interview loop, and the opportunity cost of the right person in the right seat (multiplied by the number of times you had to try to fix it).
3. Read an example (or a few). The a16z hiring process doc [11] is terrific. Our tool, the MOC, is crafted just for this purpose. If you’re truly patient, read WHO: The A Method for Hiring [12]. Ask a friend at a company with 20-50 or 50-100 employees to see how they do this.
4. Build your first MOC. On your next hire, build an MOC before the first interview. Full stop. At five or ten people, this is a one-page, two-hour artifact whose whole job is forcing you to articulate the outcomes and the one strength that matters. The full process — interviewer slices, mechanical scoring — can come once the team has grown past ~20. And hey, since I know you’re already groaning or preparing to avoid doing it, use an AI to help you. Our tool version of the MOC is built to make that process painless. If you use AI without our prompts, make sure not to outsource all the thinking by asking it to do it for you. Have it interview you.
5. Define your team’s roles. Role definition isn’t just for recruiting & hiring. You can also define the roles for each person on your team. They’ll thank you for it, and it will be good practice for when you are hiring. If defining eight roles sounds daunting, just pick one.
6. Wire it into your hiring process. At five people, I’d be shocked if you had a hiring process. At twenty, it’s probably held up with duct tape. At fifty, you’ll have one — because if you don’t, you’ll never get to fifty. But the more you wire in a real process for recruiting and hiring your team, the more you’ll find you recruit great talent. When you make mistakes, go back and correct the process.
7. Calibrate, calibrate, calibrate. If you’re managing a team, especially if you’re an executive, and especially if you’re a CEO, you’re going to be doing this a lot. And you’re going to be hiring people for roles you had no idea existed, let alone how they work. The answer to, “How do I know how to hire X?” is deceptively simple: run ten calibration calls with people who’ve done the role well, hired for it successfully, and worked alongside someone great at it. If you’ve hired fifty people into this role across your career, one or two might do. But don’t stop. The world changes, and it’s happy to change without you.
8. Buck up. This is a lot of work. But you know what’s also a lot of work? Working with coworkers you don’t get along with and having to fire people. If this stuff sounds unappetizing and you’d rather drink mai tais on the remote work beach in Bali, get a different job.

It is not uncommon for a salary at a startup scaling from 1 to 100 employees to exceed $200-300K for senior people. If you plan on them staying for four years, you’re staring at a ~$1M decision in cash alone. To account for the value the right person can create and their impact up, down, and across, you should think about a senior hire as at least a $2M investment.

If you were investing $2M, or even buying something for $2M, like, say, a house, you would conduct a proper bit of research. You’d look up the schools, fight about which directions the windows face, and you’d pay a professional inspector before signing. For the multi-million dollar hire, many founders spend a couple hours, borrow a job description, post it, and call it a day. That, my friends, is what a group of researchers at Penn call cognitive surrender [13]: handing your thinking to an external artifact. Their research is about surrendering to AI, but I’ll extend it. Whether you surrender to a job description posted on HackerNews or to an AI makes no difference. Surrender is surrender and it ain’t how legacy-building companies get built.

Like it or not, you’re going to have to do the thinking. And sometimes, counterintuitively, that means using artificial intelligence to augment your own. Use our prompts to set up a session with ChatGPT or Claude to interview you to create a great role definition document. Read the document. Edit it. And you know what, you may have to write it in your own language after it’s done. The same Penn research explains why the reading and editing are not optional: AI help makes you smarter when the model is right and drags you below your own unaided baseline when it’s confidently (and boy is it confident) wrong [13]. Once you’re comfortable, use these tools offensively: give an AI the rubric and the interviewers’ written scores and let it add up the mechanical total — arithmetic, not vibes [2] — then invite it back to red-team the finished document. Whether you want to do this or not, know that someone will. And that’s your competition.

Other failure modes, in roughly the order I see them:

1. Consensus definition: One person must own the definition process. It is their responsibility to collect input from everyone who will interact with the role and decide what to integrate.
2. Right person, wrong terrain: Many people who perform great at a big company will burn out quickly (and suffer) at a startup. Use the Terrain Test to match a person’s stage-relevant skills to yours.
3. Ignoring widow-makers: Peter Drucker, the famous management theorist, called jobs that defeated two proven performers in a row “widow-makers,” after ships that kept killing sailors until the owners stopped blaming the crews and decided to retire these murderous vessels [6]. When two people who’ve succeeded elsewhere fail in a role you’ve defined, the role needs to change. And Drucker’s corollary matters as much as the rule: the answer is never to go hunting for a universal genius who can survive the role as written. A job ordinary competent people can’t perform is a job you can’t staff, so restructure the assignment [6].
4. Ignoring your gut: When you, the hiring manager, have a deep conviction in something, whether that’s a need for the team or a yucky feeling you got when you entered the room with someone, it is your job to make that thing legible and advocate for its importance as a filter. Don’t hire people when your gut screams no. Don’t do it twice. You still will. But pay attention to how you feel when you enter and exit the room with people. Treat your gut like a smoke detector, not a judge. When it screams, stop and figure out how to make the signal legible. Identify the behavior, add it to the rubric, and then use the rubric to vote. What your gut does not get is a veto after the scores are in; that override is exactly how the research shows humans fail [2].
5. Hiring shoulds instead of needs: You’ve been told you should hire X. Deep down, you think you need a Y. Too many leaders talk themselves out of their own judgment, blaming inexperience or functional naivete.
6. Failing to block time: A few hours on your calendar this week. And next. And each week you’re hiring for new roles.
7. Inflating requirements: Some teams have a habit of piling every nice-to-have into a role definition document. And then they’re surprised when nobody alive matches it. In one Harvard survey, almost nine in ten employers admitted their screening software filters out qualified people for missing criteria the job doesn’t actually need [14]. Keep the bar high for outcomes and the required competencies reasonably short. Write the must-haves, name the acceptable weaknesses, and remember nobody’s perfect.

Yes, defining roles can be a lot of work. And if you’re defining roles as laid out here and still struggling, something else is probably broken. But the sheer volume of complaints I’ve seen from leaders about their teams compared to the excuses they make for skipping the work is like choosing to go to Olive Garden and being upset it’s not Flour + Water. Extraordinary outcomes are the result of extraordinary effort, and as the CEO, founder, leader, or manager, decide if you want to be a mid-range pasta factory or an award winning restaurant.