Human capital trades more like real estate than equities. While millions in S&P 500 stocks can exchange hands with a single click, hiring employees or finding life partners always requires a final offline, high-touch evaluation - much like inspecting a house before closing the deal.
In the job market, some predictive attributes can be translated into text - whether they went to college, what they studied, etc. But for several things we care about - integrity, confidence, ability to think on one's feet, and most importantly, red flags that are quite hard to put a finger on, a 15-minute in-person interaction can tell us things that resumes, cover letters, or a thousand other fields on a digital form can't. Much of our cognitive machinery has evolved precisely to observe and evaluate other human beings, and to spot deception.However, these in-person interactions or even phone calls are much less scalable. In any evaluation process, you can do a fixed number of them. The typical recruitment funnel is set up to maximize the expected value of these high-touch, personalized interactions.
The earliest stages of the funnel typically involve a low-touch process like screening resumes where recruiters look for signals that might correlate with success (being hired). But at this stage, the type of information that's available is very different. You get almost no information about things like trustworthiness but get data points on degrees, achievements, experience, etc. An ideal recruiter is basically trying to identify the most predictive fields out of those available to select people for the next stage of the process, even if these fields are not necessarily the primary determinants of success in the final stage.
Example
Let's say XYZ is a think tank looking to hire a Research Analyst. XYZ is a small organization working on fiscal policy. They are looking for someone that combines quantitative skills with written communication skills. Some research experience is a must. Experience in fiscal policy is not a must. Since XYZ is lean, it's important that the new hire be able to work independently with little supervision and handle uncertainty and ambiguity.
For simplicity, assume they only have two stages to their evaluation process. The first is a resume and cover letter screen, and the second involves in-person interviews and a short research assignment. Given how time-consuming the second phase of the selection process is, XYZ has decided they can only invite 5 applicants to this stage.
The CEO of XYZ has laid out the following rubric for evaluating the new hire:
The recruiter from XYZ receives 1,000 resumes and cover letters. They decide they will screen all the resumes (spending <1 min per resume) to select 100 applicants for further screening. They will read cover letters for these 100 applicants before deciding which 10 applicants they invite to the second round interviews.
When the recruiter screens the resumes, the only metric they can reliably screen for is familiarity with fiscal policy - since this would manifest in prior work experience, internships, or coursework. The recruiter can also check if applicants have experience that might be indicative of the ability to handle quantitative work. One could perhaps make a weaker case for written communication skills, but with written communication, the question is one of quality, not whether someone can literally write. Moreover, almost everyone is likely to include something that claims they've written research reports. Even though relevance of work experience is far from the only metric that matters, it's the only one in this case that can be reliably screened for.
The disadvantage faced by unconventional applicants
Now imagine you were applying to this job. You have strong quantitative skills and have been writing on Substack - mostly data-heavy pieces on areas like immigration policy. You've gotten good feedback and grown your subscriber base by yourself. This already makes you a pretty good fit for the job. You've demonstrated that you can work independently and have validated your research and quantitative skills without any institutional support. You also have a math major, so working in a different quantitative domain is unlikely to be a problem for you. But if the recruiter were rating you (out of 10) on relevance of work experience, they might give you a 6 or 7, especially if a majority of their applicants have worked in other think tanks, for example.
Now, of course, if you could make the recruiter click on your Substack or even chat with you for 2 minutes, you would probably be invited to the second round. But the recruiter doesn't have time to validate this; they only have time to eyeball the relevant metric on which they can rank people into 0s and 1s. From the perspective of the recruiter, even if they're well-informed about what matters on the job, their immediate goal is to find a way to rank order the large pile they have in front of them, and on that metric, you're not very likely to make it through to the next round.
The role of institutional incentives in recruiter behavior
Institutional incentives probably matter too. Recruiters are probably not incentivized sufficiently to maximize the expected value of the hire. If they are in-house, they are probably paid a salary to do a job, in which case they will care first and foremost about not being perceived as incompetent. If this Substack writer guy turns out to be a great hire, the recruiter won't get applauded. But if he presents as a moron during the interviews, then the CEO might ask the recruiter why they brought someone who had little relevant work experience.
Sometimes firms hire external recruiters who are compensated in the form of commissions if the firm hires a candidate they bring in. In this case, recruiters do have an incentive to bring in people that get hired. But they are still going to try to sort by fields that are most legible in the early stages. They don't have the bandwidth to go find gems in the rough.
This presents a challenge for a few types of applicants - those who are looking to transition to a new area/industry, those who are a good fit because they have done something analogous in a different setting, and applicants who seem average on legible metrics but exceptional on soft skills. Basically, if you meet every item on the preferred qualifications list on a job application, you might be able to get away with applying on job portals, but otherwise, you're at a significant disadvantage. And so is the firm. But this is how the world works. Feedback loops are not very strong, and companies don't actually know if they end up with sub-optimal applicants long after they hire someone, and even then, they don't know the counterfactual of how much they would have accomplished with an even better candidate.
Startups and Advserse Selection
On the employer side, this is particularly bad for startups and other smaller organizations with low brand recongition/prestige.
For startups and growing companies (10-200 employees), traditional hiring channels create a particular challenge: when posting jobs on LinkedIn or similar platforms, startups must keep the application process simple to receive any applications at all - they lack the brand recognition of larger companies. But when the cost of applying approaches zero, they face a deluge of applications, many from candidates who apply indiscriminately to anything remotely relevant.
This creates two compounding problems:
The applicant pool becomes disproportionately filled with "spray and pray" candidates - those applying to everything without careful consideration of fit
To manage the volume, recruiters must screen on highly legible criteria like specific experience or recognizable company names
The resulting candidate pool often contains just two types of people:
Those in high demand from many organizations (including larger, more established companies that can offer better compensation)
Those who might be underperformers on less visible traits that only become apparent later in the screening process
This matters particularly for startups because:
Each hire represents a larger percentage of total company output
Early employees disproportionately shape culture and long-term success
They often can't compete with larger companies on tangible benefits and compensation
Implications for Job Seekers and Startups
For job seekers, especially those with non-traditional backgrounds or transferable skills, the implication is straightforward: don't spend all your time applying to jobs. The legibility trap means that unless your resume perfectly matches the job requirements, you're likely to be filtered out regardless of your actual capabilities. While large companies have formalized employee referral networks to partially address this problem, startups - who need great talent the most - often lack the scale and systems to effectively leverage referrals.
This points to a deeper challenge in the job market: the information about who might be great at a role - despite not having the most conventional resume - exists distributed across professional and social networks. A former colleague might know someone who built something impressive at a small company, or a mentor might know someone looking for exactly the kind of challenge your startup offers. But without proper incentives and systems to activate these networks, this valuable information remains dormant.
Looking Ahead
In my next post, I'll explore the incentive structures that make referral networks work (or fail) and ideas for systems that encourage high-integrity referrals by aligning incentives.
I'm working on testing minimum viable solutions to this problem at the moment. If you're interested in this space - whether as a potential user, hiring organization, collaborator, or someone with relevant insights - you can reach me at vaishnav@probablygood.org.
Human capital trades more like real estate than equities. While millions in S&P 500 stocks can exchange hands with a single click, hiring employees or finding life partners always requires a final offline, high-touch evaluation - much like inspecting a house before closing the deal.
In the job market, some predictive attributes can be translated into text - whether they went to college, what they studied, etc. But for several things we care about - integrity, confidence, ability to think on one's feet, and most importantly, red flags that are quite hard to put a finger on, a 15-minute in-person interaction can tell us things that resumes, cover letters, or a thousand other fields on a digital form can't. Much of our cognitive machinery has evolved precisely to observe and evaluate other human beings, and to spot deception.However, these in-person interactions or even phone calls are much less scalable. In any evaluation process, you can do a fixed number of them. The typical recruitment funnel is set up to maximize the expected value of these high-touch, personalized interactions.
The earliest stages of the funnel typically involve a low-touch process like screening resumes where recruiters look for signals that might correlate with success (being hired). But at this stage, the type of information that's available is very different. You get almost no information about things like trustworthiness but get data points on degrees, achievements, experience, etc. An ideal recruiter is basically trying to identify the most predictive fields out of those available to select people for the next stage of the process, even if these fields are not necessarily the primary determinants of success in the final stage.
Example
Let's say XYZ is a think tank looking to hire a Research Analyst. XYZ is a small organization working on fiscal policy. They are looking for someone that combines quantitative skills with written communication skills. Some research experience is a must. Experience in fiscal policy is not a must. Since XYZ is lean, it's important that the new hire be able to work independently with little supervision and handle uncertainty and ambiguity.
For simplicity, assume they only have two stages to their evaluation process. The first is a resume and cover letter screen, and the second involves in-person interviews and a short research assignment. Given how time-consuming the second phase of the selection process is, XYZ has decided they can only invite 5 applicants to this stage.
The CEO of XYZ has laid out the following rubric for evaluating the new hire:
The recruiter from XYZ receives 1,000 resumes and cover letters. They decide they will screen all the resumes (spending <1 min per resume) to select 100 applicants for further screening. They will read cover letters for these 100 applicants before deciding which 10 applicants they invite to the second round interviews.
When the recruiter screens the resumes, the only metric they can reliably screen for is familiarity with fiscal policy - since this would manifest in prior work experience, internships, or coursework. The recruiter can also check if applicants have experience that might be indicative of the ability to handle quantitative work. One could perhaps make a weaker case for written communication skills, but with written communication, the question is one of quality, not whether someone can literally write. Moreover, almost everyone is likely to include something that claims they've written research reports. Even though relevance of work experience is far from the only metric that matters, it's the only one in this case that can be reliably screened for.
The disadvantage faced by unconventional applicants
Now imagine you were applying to this job. You have strong quantitative skills and have been writing on Substack - mostly data-heavy pieces on areas like immigration policy. You've gotten good feedback and grown your subscriber base by yourself. This already makes you a pretty good fit for the job. You've demonstrated that you can work independently and have validated your research and quantitative skills without any institutional support. You also have a math major, so working in a different quantitative domain is unlikely to be a problem for you. But if the recruiter were rating you (out of 10) on relevance of work experience, they might give you a 6 or 7, especially if a majority of their applicants have worked in other think tanks, for example.
Now, of course, if you could make the recruiter click on your Substack or even chat with you for 2 minutes, you would probably be invited to the second round. But the recruiter doesn't have time to validate this; they only have time to eyeball the relevant metric on which they can rank people into 0s and 1s. From the perspective of the recruiter, even if they're well-informed about what matters on the job, their immediate goal is to find a way to rank order the large pile they have in front of them, and on that metric, you're not very likely to make it through to the next round.
The role of institutional incentives in recruiter behavior
Institutional incentives probably matter too. Recruiters are probably not incentivized sufficiently to maximize the expected value of the hire. If they are in-house, they are probably paid a salary to do a job, in which case they will care first and foremost about not being perceived as incompetent. If this Substack writer guy turns out to be a great hire, the recruiter won't get applauded. But if he presents as a moron during the interviews, then the CEO might ask the recruiter why they brought someone who had little relevant work experience.
Sometimes firms hire external recruiters who are compensated in the form of commissions if the firm hires a candidate they bring in. In this case, recruiters do have an incentive to bring in people that get hired. But they are still going to try to sort by fields that are most legible in the early stages. They don't have the bandwidth to go find gems in the rough.
This presents a challenge for a few types of applicants - those who are looking to transition to a new area/industry, those who are a good fit because they have done something analogous in a different setting, and applicants who seem average on legible metrics but exceptional on soft skills. Basically, if you meet every item on the preferred qualifications list on a job application, you might be able to get away with applying on job portals, but otherwise, you're at a significant disadvantage. And so is the firm. But this is how the world works. Feedback loops are not very strong, and companies don't actually know if they end up with sub-optimal applicants long after they hire someone, and even then, they don't know the counterfactual of how much they would have accomplished with an even better candidate.
Startups and Advserse Selection
On the employer side, this is particularly bad for startups and other smaller organizations with low brand recongition/prestige.
For startups and growing companies (10-200 employees), traditional hiring channels create a particular challenge: when posting jobs on LinkedIn or similar platforms, startups must keep the application process simple to receive any applications at all - they lack the brand recognition of larger companies. But when the cost of applying approaches zero, they face a deluge of applications, many from candidates who apply indiscriminately to anything remotely relevant.
This creates two compounding problems:
The resulting candidate pool often contains just two types of people:
This matters particularly for startups because:
Implications for Job Seekers and Startups
For job seekers, especially those with non-traditional backgrounds or transferable skills, the implication is straightforward: don't spend all your time applying to jobs. The legibility trap means that unless your resume perfectly matches the job requirements, you're likely to be filtered out regardless of your actual capabilities. While large companies have formalized employee referral networks to partially address this problem, startups - who need great talent the most - often lack the scale and systems to effectively leverage referrals.
This points to a deeper challenge in the job market: the information about who might be great at a role - despite not having the most conventional resume - exists distributed across professional and social networks. A former colleague might know someone who built something impressive at a small company, or a mentor might know someone looking for exactly the kind of challenge your startup offers. But without proper incentives and systems to activate these networks, this valuable information remains dormant.
Looking Ahead
In my next post, I'll explore the incentive structures that make referral networks work (or fail) and ideas for systems that encourage high-integrity referrals by aligning incentives.
I'm working on testing minimum viable solutions to this problem at the moment. If you're interested in this space - whether as a potential user, hiring organization, collaborator, or someone with relevant insights - you can reach me at vaishnav@probablygood.org.