Why do qualified people get filtered out before anyone reads their resume?

By Joseph Braithwaite, founder of MPE · First published October 23, 2026

The market isn’t what’s keeping Jack Weis out. We are.

Why a Princeton PhD can’t get hired, and what both sides of hiring need to change

Everyone who read the Wall Street Journal piece on Jack Weis landed in the same place. He has a Princeton PhD in chemical engineering and research the Nobel committee cited, he’s sent more than 2,300 applications in two years and has nothing to show for it, and the explanation that came with the story was that it’s a frozen market, with low hire and low fire rates, meaning nobody is leaving so nobody is getting in.

I don’t think that’s the whole story.

The freeze is real, Bruce Fallick, an economist at the Federal Reserve Bank of Cleveland, published research on it this summer. A frozen market limits how many jobs there are, but it doesn’t explain why not one person reading those 2,300 applications could find a way to say yes to him, right down to a food-prep job at a company that sells frozen crab cakes.

That part is on us, the people deciding who gets through, and I’m including myself. I’ve spent 25 years in enterprise transformation and I’m now at 5,003 applications of my own with a response rate of about 1%, so I know how this looks, I’m living it too.

None of this is news either. Harvard Business School and Accenture put numbers on it back in 2021 with their Hidden Workers research, and 88% of the employers they surveyed agreed their own screening was turning away qualified, high-skilled candidates because they didn’t match the exact criteria (keywords) in the job description. Five years later we all know it and we still screen candidates the same way.

What do you see when a PhD resume lands in front of you?

You see “PhD, chemical engineering” and file it under research and academic papers - someone who’ll be brilliant at the one narrow thing they studied and lost anywhere else - and you move on to a resume that looks more like what you believe the job needs. What you skipped past is six years working a problem nobody had solved, building a way to test an answer, watching it fail and working out why, then standing in front of people whose whole job was to tear the result apart and defending it until it held. That’s problem solving and solution validation done harder than most of us will ever have to do it, and it’s what most of the senior roles you recruit for are asking for, the postings just call it “drives to root cause” or “evidence-based decision making” so nobody connects the two.

Jack Welch had the same degree, a chemical engineering PhD from Illinois in 1960, and he ran GE for twenty years. Think what you want about his leadership style, but his results running a global behemoth that sold everything from jet engines and MRI machines to mortgages speak to his ability to connect the dots, solve the hard problems and weigh the details, all things he would have picked up getting his PhD. GE gave him a first job close to what he’d studied and he proved the rest from the inside, where people could watch him do it. Somewhere along the way, we removed those opportunities; now people like Weis have to prove all of it from the label on their resume before anyone lets them past the reception desk.

Are you really hiring the best person?

You’re not, you’re hiring the closest copy of whoever sat in the seat last. Job postings all get written around that person, their title and the words they used for the work, not the problem the role is meant to solve. Peter Cappelli at Wharton has been making this point for years in his research on hiring, that job descriptions turn into a wish list modelled on the employee who just left. Then the ATS ranks the applications against those words, on criteria the recruiter set up, and the recruiter takes the top of that list and checks titles and logos. The newer systems call it skills matching or semantic search, but it’s still the posting’s words compared against the resume’s words, just with more words counted as ‘close enough’, and nothing in that score says whether the person has done the work successfully. Anyone who described the same work in research language, or different nomenclature ends up so low on the ATS pre-filtered list that nobody sees them. Most of Weis’s applications never got an answer at all, not even a no, and with hundreds of applications on a req the likeliest reason is he didn’t rank high enough and nobody read far enough down to get to him.

And 2,300 applications aren’t 2,300 separate chances either. A Stanford team led by Rishi Bommasani studied four million applications that employers had screened with algorithms from one vendor, and found certain applicants marked for rejection on job after job, more often than chance would explain. They call it “algorithmic monoculture”. The screening software behind different employers’ systems is so similar that the results come out more alike than chance would explain, so a resume that ranks low at one company is more likely to rank low at the next one using the same or similar technology.

How do I know it’s us?

I don’t know for sure, and there are other things that could explain some of it.

The research market he was aiming at did get worse. Indeed’s Hiring Lab found science and R&D job postings dropped 18% in the first five months of 2025, compared to 4% for jobs overall, after the US federal research cuts, and that hit in his first year of looking. It’s also possible something in his applications was putting people off. His mother read through them looking for a problem and didn’t find one, but I haven’t seen them and neither has anyone else writing about him.

The other explanation I keep hearing from recruiters is that he’s overqualified and would leave as soon as something better came along - we haven’t had lifelong jobs for a few decades now, but this theory of terminal devotion to a company still hangs on as an unspoken requirement. He might leave, but so might anyone you hire at any level of education. Somebody decided what a PhD would or wouldn’t put up with based on the degree, which is the same thing as reading “PhD” and seeing “academic,” and that was a person making a hiring decision.

The research cuts would explain why the lab jobs dried up, and software hiring has been tight too, but a food-prep job at a crab cake company doesn’t depend on research funding and they didn’t answer him either. I can’t prove it was how his applications were read, or not read. Nobody can prove it wasn’t, because no employer goes back to check whether the PhD they skipped could have done the job, and after 2,300 applications that’s worth asking about.

Isn’t that what your judgement is for?

That’s the defence I hear most, that the software only sorts and a person with experience makes the call. A federal court in California has already rejected the first half of it. In Mobley v. Workday the judge ruled this March that scoring and ranking applicants is part of the hiring decision, so the vendor that does it for an employer can be held liable as that employer’s agent. The second half assumes your judgement is the same every time you use it, and it isn’t.

Kahneman, Sibony and Sunstein wrote a whole book on this -“Noise”, showing the same professional reading the same file reaches a different verdict depending on the day and the hour, so the resume you read at 9 on Monday and the one you read at 4:30 on Friday aren’t held to the same bar and you’d never know it. Kahneman also spent years working through when gut calls can be trusted at all with Gary Klein, who’d built his career defending expert intuition, and in 2009 the two of them agreed on two conditions. The situation has to be regular enough to learn from, and you have to get quick, clear feedback on whether you were right. Recruiting fails the second one completely, because you never find out how the person you passed over would have done, so your sense of who’s good never gets corrected and it just keeps feeling right with no validation or evidence.

There’s another problem that’s harder to spot. Even on your best morning, how you read a resume is shaped by where you’ve been. In a study published in 1954, the psychologists Albert Hastorf and Hadley Cantril showed Princeton and Dartmouth students the same film of a rough football game between the two schools, and the Princeton students counted roughly twice as many Dartmouth fouls as the Dartmouth students did. Lauren Rivera found the same thing inside hiring when she spent years studying how elite banks, law firms and consulting firms pick new hires, where evaluators kept choosing people who reminded them of themselves, people who’d come through the same kind of schools and spent their weekends the same way, and calling it fit. If you’ve never done a PhD you see papers and miss the problem solving, and ten seconds isn’t long enough to see anything else but a logo and your own lack of a PhD.

What you call sound judgement is an opinion built from your own history, which works fine right up until you have to defend it for someone whose background is nothing like anyone you’ve hired before. Jennifer Lerner and Philip Tetlock pulled together decades of research on accountability in 1999, and it shows that when people know they’ll have to answer for a decision they lean toward the choice with the most obvious reasons behind it. So you pick the candidate you can explain, the one more like you.

Why does every req take so long?

This is where the logjam everyone complains about comes from. The posting is written for the person who left, the ATS ranks the closest copies of that person at the top on criteria the recruiter set up, and the recruiter shortlists whoever looks most familiar from the top of that list. But nobody on the list is the person who left, so nobody fits the posting perfectly, and anyone who’s a bit different is a risk the recruiter or the hiring manager has to put their name to. So the decision gets pushed to another interview round and another opinion, and the job sits open for months while the client is told the market is tight, and the people who could have done the job are still near the bottom of the list where nobody read them.

The usual fixes don’t touch it. Another interview round is another opinion on the same shortlist, and decades of interview research shows two people watching the same candidate often don’t agree and tend to make up their minds early, then spend the rest of the conversation confirming their opinion. Networking used to be the way to get your resume read, but people remember what you can do for them afterwards, not the value you could bring, so it mostly works for whoever is already inside - Weis called his Princeton and Rice alumni and was usually told to check the company’s job board again.

So what actually changes it?

Stop hunting for a perfect match to a posting written for the last person in the seat, and standardise the first evaluation. That means both sides doing something different. On the hiring side, the hiring manager has to write out what the role actually needs done, the problems it has to solve and at what level, instead of reusing the last posting and the last person’s title, and the recruiter has to read every resume against that for evidence the person has done that kind of work and at what level, the same way for the first candidate on Monday morning and the four hundredth on Friday afternoon. Your judgement then goes where it’s actually worth something, on the interviews and the hiring decision, working from candidates who were all measured against one standard.

On the candidate side, you still have to write your resume to get through the ATS, so use the posting’s words where they’re true to your work, but every line under them has to scream value. Write it in CAR format, the Challenge you faced, the Action you took and the Result you got, in plain words someone outside your field can follow. Weis can keep the PhD at the top, but under it he needs lines that show the challenge he was handed, what he did to find out why it failed and what held up when he proved the fix, because that’s the work most senior roles are asking for. That doesn’t let the hiring side off the hook, a clear resume still has to be read by someone, but it makes the evidence easier to find when they do.

Standardising that first evaluation is the step I built MPE to do. It reads the job description for the capabilities the role asks for and checks each resume for evidence of them, quoting the lines behind every score so you can check it. It doesn’t make the hire and it doesn’t replace your judgement, and it can’t fix junk going in, a posting copied from the last person or a resume full of job duties with no results gives it nothing to measure.

Which req on your desk right now could someone like Jack Weis have filled?

Sources: Wall Street Journal on Jack Weis (via Entrepreneur) · Lindsay Ellis · Cleveland Fed, The “Low-Hire, Low-Fire” Labor Market (Fallick, 2026) · Harvard Business School and Accenture, Hidden Workers (2021) · Bommasani et al., Algorithmic Monocultures in Hiring (Stanford, FAccT 2026) · Indeed Hiring Lab via Bloomberg, science R&D postings (May 2025) · Mobley v. Workday (N.D. Cal.) · Kahneman, Sibony and Sunstein, Noise (2021) · Kahneman and Klein, Conditions for Intuitive Expertise (2009) · Hastorf and Cantril, They Saw a Game (1954) · Rivera, Pedigree (2015) · Lerner and Tetlock, Accounting for the Effects of Accountability (1999) · Cappelli, Wharton

First published on LinkedIn, October 23, 2026. Related: Does the ATS auto-reject your resume?