Every résumé looks great on paper. Here's what the job market quietly does to the stack, and what it means for you.
Résumés are written to impress. On average, people deliver about what theirs promise. But most fall a little short, and a few are far better than they look.
Out of every 100 people applying to your job post, how many promise more on their résumé than they deliver?
Out of every 100 people applying for the same jobs as you, how many promise more on their résumé than they deliver?
Tap one to jump to the chapter that answers it, or read straight through.
The people applying today aren't a random slice of everyone. Watch how the mix shifts over a few years.
Same headline, same keywords, same confident bullet points. From the outside, you can't tell them apart.
Each tile moves to what that person actually delivers compared with what their résumé promised. 1× means exactly as advertised.
come in under their résumé. The average still lands right on 1×.
Getting hired runs on résumés, so everyone gets offers at about the same pace.
Staying runs on fit. In this model, the further a role and a person are from what the résumé promised, the sooner they part ways.
People move in and out of jobs. Matches that work tend to last. Mismatches tend to end, and those people are back on the market more often.
Scroll up to rewind the clock.
of the people applying at any moment come in under their résumé, up from 59% overall. Nobody became less honest. The market just keeps re-sorting.
On average they deliver 83 cents per résumé rupee.
Only about 8% of people are on the market at any moment, and a public job post reaches them first. Plenty of them are excellent. The stack as a whole is just skewed, so every résumé gets less benefit of the doubt.
Recruiters see the whole stack, not you. Because the stack leans toward over-promising, every applicant gets less benefit of the doubt, including the great ones. The suspicion you feel is about the pile you're in.
41% of people deliver more than their résumé says. They're on the market about half as often as everyone else (5% vs 10%), so a résumé screen is the worst place to find them.
There's a 41% chance your résumé undersells you. People like that are on the market half as often (5% vs 10%), so companies rarely see enough of them to notice.
A model keeps one reason people part ways (fit) to show the effect clearly. In real life people leave for many reasons, like layoffs, bad managers or moving cities, and many people on the market are brilliant.
A great conversation tells you less about a person than it feels like it does.
r is a correlation: one number from 0 to 1 for how well one thing predicts another. Here, how well an interview score predicts how someone actually performs once hired.
of the people your interview ranks in its top 20% are truly top-20% performers. Pure luck would give 20%.
How it's measured: researchers interview people, track how they actually perform on the job later, then check how closely the two line up. The figures here (.18 for a casual chat, .44 for a structured interview) come from a 2022 review of decades of these studies by Sackett, Zhang, Berry & Lievens.
Before you meet a candidate, all you know is the stack they came from. What they'll deliver could be anywhere under this curve.
Before they meet you, an interviewer only knows the stack you came from. To them, you could be anywhere under this curve.
80% of people on the market sit between 0.42× and 1.33×.
It goes well. But research puts an unstructured interview's predictive power at just .
That's how much the range shrinks.
The same questions for everyone, scored on a rubric: , the best-validated format there is.
narrower. Most of the blur is still there.
At , better than any method ever measured, the range only shrinks by about half.
Accuracy pays off slowly until it's close to perfect.
Meet five people with equally good résumés, using the same structured interview, and hire the top scorer. Your hire's expected delivery climbs from 0.82× to 1.01×.
Because one read is blurry, companies compare several people and pick the top score. You can do well and still lose a close race. That says more about the blur than about you.
We draw a real person from the market. The yellow line is what they truly deliver, which you'd never see in real life. Interview them a few times and watch the blur clear, slowly.
Pick how good your strongest finalist truly is and how many others they're up against. We replay that final round 100 times under each kind of interview.
Pick how good you truly are and how many others are in the final round. We replay that final round 100 times under each kind of interview.
Each square is one final round, replayed 100 times. is the candidate; whoever pops up got the offer.
One round shouldn't decide: a truly top-20% candidate scores below the average applicant in a single interview 19% of the time when it's structured, and 36% when it's a casual chat.
Off-days are normal: a truly top-20% candidate scores below the average applicant in a single interview 19% of the time when it's structured, and 36% when it's a casual chat.
The same interview gives very different results depending on where the candidates came from.
Each line meets five candidates, reads them as well as that route allows, and hires the best. Where the line ends is where that hire really ranks among everyone. Further right is better.
From a job post, judged on charm alone: the 38th percentile.
An unstructured interview reaches the 44th. A structured one reaches the 51st.
That's about as far as a job post goes, because you're ranking five people from a skewed stack.
Pick someone at random from people who aren't job hunting right now, with no interview at all: 51st. The same as the job post's best effort.
Reach out to people who aren't applying, then run the same structured interview: 66th percentile.
That's why recruiters keep calling once you have a job: someone already picked you and kept you, and they want to borrow that signal.
A friend who's job hunting: 53rd. An ex-colleague who has seen the work, vouching for someone in a job: 67th. A colleague's view predicts performance at about .
Ask a top-10% employee: “Who's the best person you've ever worked with?”
Stay in touch with people who have seen your best work. When they're asked “Who's the best person you've worked with?”, your name skips the stack.
if their network is only loosely like them (). Keep scrolling to tighten it.
Nobody has measured how closely a strong performer's contacts match their quality. Slide it and watch where their referral's line ends.
From the source model: ρ = 0 → 67th, .3 → 77th, .5 → 82nd, .8 → 88th percentile. Values between are interpolated.
When an employer fights to keep someone, the people who know their work best are telling you something.
They already have a job. You make an offer. Now their employer gets a say.
You already have a job and another company makes you an offer. Now your employer gets a say.
of people with an outside offer get a counter-offer.
Years of working with someone gives an employer a read of about , far sharper than any interview.
Each block is a quarter of an average person's output. The people they fight for deliver 1.75×. Everyone else: 0.91×.
An employer only lets someone go when they value them less than you do. Win half the fights, and those hires still deliver 1.48×, against 0.91× for someone nobody competed for.
more output from the same résumé.
If an average engineer earns ₹30 lakh and creates ₹60 lakh of value a year, that gap is worth about ₹34 lakh a year. Paying a premium up front beats losing them to a counter.
Pay bands hide how good you are, and the clearest proof shows up when someone tries to hire you away. Whatever you decide, a counter tells you how much you're valued.
Chance that someone's current employer counters your offer, by how good they truly are. The better the person, the more likely you'll need to win a second round.
Chance your current employer counters an outside offer, by how good you truly are. Employers fight hardest for the people they rate highest.
Each room holds 100 people with an outside offer. A raised hand means their employer made a counter-offer.
A countered hire you win delivers 0.57 of an average person's output more than someone nobody competed for. Put in your own numbers.
Candidates show up one at a time, and they don't wait. When does a company say yes?
Each stick is one candidate, left to right in the order they turn up.
The middle line is an average candidate. A stick pointing up is better than average; pointing down, worse. The longer the stick, the bigger the gap.
Colour shows how good they truly are: purple is the top 20%, orange the bottom 20%. For now, pretend you can see this.
Let the first 37% pass without hiring (the shaded area). The dashed line marks the best of them. Then hire the first stick that rises above it, circled.
It's the classic secretary problem. It finds the single best candidate about 36% of the time.
If a top-5% hire is good enough, only benchmark the first 20%.
chance of a top-5% hire, against 60% with the 37% rule, and you stop after about 53 people instead of 74.
Companies only see interview scores. The sticks now show how well each person interviewed: up means better than average in the interview.
Colours still show who's really good. Watch some purple sticks drop and some orange ones rise.
chance of a top-5% hire with the same 20% rule.
With a realistic ten interviews, a structured read finds a top-10% hire 18% of the time. A perfect read manages 30%, because ten people rarely include a star. Spend effort on who you meet, not on when you stop.
If you're among the first few people interviewed, you may be the benchmark the company uses to judge everyone after you. A “no” early in a process is often about timing, not about you.
Ten candidates, a benchmark of the first two, then hire the first who beats it. Pick a chair: the dial shows the same candidate's chance of getting the job, depending only on when they're interviewed.
A company sees ten people, uses the first two as a benchmark, then hires the first who beats it. Pick a chair: the dial shows your chance of getting the job, depending only on when you're interviewed.
Tap a chair to sit there. The first two chairs are the benchmark.
Twelve candidates, one structured interview each. You see each score as they arrive. Hire them, or let them go for good.
Everything in one line each. Tap a card to flip it.
Applicants to a public post lean toward résumés that over-promise, through no fault of the great people in it. Judge everyone with the same structured method.
Use structured interviews, run several with different interviewers, and always compare candidates side by side.
Ask “Who's the best person you've worked with?”, not “Know anyone looking?” It reaches great people who would never see your post.
A counter-offer is a sign you picked well. If they counter, counter back.
Use the first 10–20% of the people you expect to meet as your benchmark and hire the first who beats them. If nobody does, go back to the best: it lifts a strong early candidate's chances from 0% to 39%.
Applications from a job board are read with suspicion because of the pile, not because of you. Plenty of great people are in it.
Work samples, portfolios, take-home tasks and structured interviews give a clearer read than a résumé ever can. Seek them out.
Ex-colleagues and managers are the strongest route in. Keep in touch, and ask them to put your name forward.
If your employer fights to keep you, that's a strong signal of your value. Use it to understand your worth, whichever way you go.
Interviews are blurry and timing matters. Even a top-10% candidate wins a five-person final only about 41% of the time, so expect around 2–3 final rounds per offer. Early slots are often the benchmark. Keep going.
Structured interviews are the best-validated format there is, and they're what we run at Intervue: the same questions for every candidate, scored on a rubric, by experienced engineers.
Intervue runs structured interviews: the same questions for every candidate, scored on a rubric by experienced engineers. Fair to everyone, whatever their résumé says.
See how Intervue works ↗