It takes about 24 job applications to get an interview if you rewrite your resume for each posting, and about 48 if you send the same resume every time.
Those two numbers come from 139,927 applications, and they are the most useful thing anybody can tell you about a job search, because they turn an open-ended activity into a rate. A rate can be measured, compared against your own, and improved. "Keep applying and stay positive" cannot.
Where the interview rate figures come from
Huntr, a job application tracker, publishes the per-application interview rates for every application its users log. Its most recent quarter covers 139,927 applications from 25,635 people, including 39,184 resumes rewritten for a specific posting.
Applications sent with a rewritten resume produced an interview 4.23% of the time. Applications sent with the same base resume produced one 2.07% of the time. Invert those and you get the two numbers at the top: one interview per 23.6 applications, or one per 48.3.
Three things are wrong with that.
It is a vendor's own data. Huntr sells the tool that does the rewriting, and "tailored" in their dataset means the applicant used that tool rather than that they tailored the resume by any other method. A company reporting that its own product doubles your results is reporting a sales figure as well as a finding.
The people in it are not average job seekers. Everybody in the sample was organised enough to log applications in a tracker. That group finds work faster than the group that applies from the sofa and cannot remember where they applied, so both rates are almost certainly better than yours.
Tailoring is not randomly assigned. Nobody rewrote a resume for a job they did not want. The applications that got the rewrite were the ones the applicant cared about, which means the 2.04 times difference measures effort and interest as well as the document. The resume is doing some of that work and not all of it.
None of that makes the numbers useless. It makes them a floor for the ratio rather than a promise about yours, and it is still the largest published sample of real applications with real outcomes attached.
Job applications rose 45% in a year while hiring did not
The reason a rate that once sounded pessimistic now sounds normal is that the number of applications went up without the number of jobs going up with it.
LinkedIn processes around 11,000 job applications a minute, a figure 45% higher than a year earlier, reported by the New York Times from LinkedIn's own data. Nothing about hiring grew 45%. What grew is the cost of applying, which generative text tools have taken close to zero.
The effect is an arms race with a predictable end. When an application costs nothing to send, people send more of them; when everybody sends more, the response rate per application falls; when the response rate falls, people send more still. The rate you are measured against is the product of that loop, not of anything you did.
It also explains why the advice to "apply to everything" produces worse results than it used to. That advice was written when sending fifty applications took a month. It now takes an afternoon, and so does everybody else's fifty.
The job market is harder than it was, and not evenly
The New York Fed asks a panel of households a question with no opinion in it: if you lost your job today, what is the chance you would find another within three months? It has asked every month for over a decade, which makes the answer a series rather than a mood.
The answer fell 4.2 percentage points in a single month to 43.1%, the lowest reading in the history of the series. Its labor market survey publishes the run and updates it monthly.
The difficulty is real rather than a failure of effort: the same people were more confident a year earlier with the same resumes. And it did not fall evenly. The decline was concentrated in households earning under $100,000, in people over 60, and in people whose education stopped at high school, which is a different market from the one most job search advice describes.
It measures expectation rather than outcome, so it reports what people believe about their own chances and not what happened to them. Over twelve years the series has tracked hiring closely enough to be treated as evidence, and it is the context for every rate below.
The job board changes the rate more than the resume does
The same dataset breaks the interview rate down by where the application was sent. The spread is larger than the tailoring effect.
| Applied through | Interview rate | Applications per interview |
|---|---|---|
| Google Jobs | 7.12% | 14 |
| 2.94% | 34 |
Fourteen against thirty-four. Rewriting a resume roughly halves the applications you need; choosing where to apply from cuts them by rather more than half.
The mechanism is not that Google Jobs is a better product. Google Jobs is an aggregator that sends you to the employer's own posting, so an application through it usually lands directly in that company's system. LinkedIn's Easy Apply lands in a queue with the other applications that were also one click, which is the queue the 11,000 a minute are joining.
The further an application travels from a one-click button, the fewer applications it competes against. Applying on the company's own careers page is the strongest version of that.
The second application to the same company does worse than the first
Splitting the applications by how many the same person had already sent to that employer produces the sharpest number in the report.
| Applications sent to that company | Interview rate | Applications per interview |
|---|---|---|
| One | 6.07% | 16 |
| Eight or more | 1.91% | 52 |
Your first application to a company is three times more likely to produce an interview than your ninth.
Part of that is a real signal being read by a real person: a recruiter looking at a candidate who has applied to eight unrelated roles sees somebody applying to a logo rather than to a job. Part of it is simple fit, since the first application is usually to the role that matched best and the eighth is to whatever was left.
Either way it argues against the common tactic of blanketing a company you like. Two well-matched applications to a company are worth more than eight, and the eight actively damage the two.
Why “fewer job applications get more interviews” is backwards
The same report shows interview rates by how many applications a person sent in total, and this is where a careless reading does real damage.
| Applications sent in total | Interview rate |
|---|---|
| 11 to 20 | 9.25% |
| 100 or more | 2.58% |
Quoted as it stands, that says sending fewer applications gets you more interviews. It is quoted that way often.
It cannot mean that. Somebody whose search ended at fifteen applications stopped at fifteen because they were hired. Somebody at a hundred and twenty is still searching, which is why they have a hundred and twenty. The two groups are not two strategies, they are the same strategy at two different points, and the successful searches leave the sample early.
The rate is measuring how long each person needed, then reporting it as though it measured how well they applied. Anyone still in the sample at 100 applications is there precisely because their rate was low.
What it does describe is the shape of a search: the ones that go quickly go quickly, and the long ones are long. That belongs in the runway calculation rather than in a decision about volume.
Nearly every large employer uses an applicant tracking system
Jobscan reverse-engineers the careers pages of every company on the Fortune 500 list and counts which ones run a detectable applicant tracking system. In its most recent count it was 487 of 500, or 97.4%.
The figure most articles quote is 98%, from an earlier count. Detectable use has drifted down rather than up, as some large employers moved to systems that do not announce themselves.
What the software does is more limited than its reputation. An applicant tracking system is a database with a search box. It stores applications, parses them into fields, and lets a recruiter filter and rank. The idea that it silently rejects three quarters of applicants against a keyword threshold describes a configuration most employers do not run, because a recruiter who never sees a qualified applicant eventually notices.
The parsing is the part that genuinely costs people interviews, and it fails on formatting rather than on content: text inside images, contact details in headers, multi-column layouts read across instead of down, skills in a graphic. A resume that parses cleanly into a plain text file will parse cleanly into an ATS.
How to work out your own interview rate
None of the published rates are yours. The point of them is to give you something to compare against, and that requires counting.
Log four things per application: the date, the company, the route you applied through, and whether the resume was rewritten for it. Then one outcome column, which is only ever "no reply", "rejected" or "interview".
Thirty applications is enough for the rate to mean something. Divide interviews by applications.
| Your rate | Applications per interview | What it means |
|---|---|---|
| Above 5% | Under 20 | The applications are working. The constraint is how many roles exist that fit you. |
| 2% to 5% | 20 to 50 | Normal. Improvements come from where you apply, not from the document. |
| Under 2% | Over 50 | Something upstream is wrong: the roles are a level off, the resume is not parsing, or the applications are all one-click. |
| Zero after 40 | Not a rate yet | Stop sending. Forty applications with no interview is not bad luck, it is a signal, and another forty will say the same thing. |
Forty applications is where the last row bites, and it is the row people carry on past. Applications are not free even when they cost nothing to send, because the hours have a price. The real hourly wage article makes the same argument about a job you already have.
How long a job search takes: 108 days is the median
The median search in the dataset ran 108 days from first application to accepted offer. Half of searches took longer.
Inside that, the waits are shorter than they feel: about six days from applying to a first response where one came at all, and about twelve days from first interview to offer. The time is not spent in process. It is spent between applications that went nowhere.
Three and a half months is a runway question, not a motivation question. If you are searching without income, the number that decides how the search goes is not your interview rate, it is how many months of spending you have. A search that has to end in six weeks is a different search: it takes the first acceptable offer, which usually means a worse one.
Two costs are routinely left out of that calculation. Health cover is the larger one, and if you left a job with it, COBRA charges the full group premium plus 2%, which for a couple in their fifties is frequently more than $1,800 a month. The article on cover before 65 covers the marketplace alternative and how the subsidy is calculated, which applies to anybody between jobs and not only to early retirees.
The second is what happens to the retirement account you left behind. It does not have to move immediately and it should not be cashed out to fund a search: the tax and the penalty together take a third of it before it reaches your account. Transferring a 401(k) from a previous employer sets out the options, and what happens if the company goes under covers the case where leaving it is not simple.
What job application data does not measure
Every figure here is about applications, because applications are what gets logged. The route that does not appear in any of it is the one that does not start with an application.
A referral does not produce a row in an application tracker until after the conversation that caused it, so no dataset built from applications can measure it. That is a gap in the measurement rather than evidence that referrals do not work. Every rate here counts one of the two ways in.
The rates also say nothing about which jobs. An interview rate of 8% for roles a level below what you did last is not a better search than 3% for roles at your level, and the arithmetic will not tell you that, because both look like interviews.
If the search is going long enough that income is the problem rather than the offer, the 25 ways to earn from home list is where the stopgap options are, and what side income is worth over time is the argument for keeping any of it once the job starts.
Frequently asked questions
How many job applications does it take to get an interview?
About 24 with a resume rewritten for the posting, and about 48 with the same resume every time, from a sample of 139,927 logged applications. Your own rate depends more on where you apply than on either figure.
How many applications a day should I send?
The dataset gives no support for a daily target. It shows that the first application to a company converts at 6.07% and the ninth at 1.91%, and that one-click applications convert at less than half the rate of applications that reach an employer's own system. Both point at fewer and better rather than at a quota.
Is it worth tailoring a resume for every job?
The measured difference is 4.23% against 2.07%, so on those figures one rewritten application is worth about two generic ones. Rewriting takes more than twice as long as sending a generic application, so the trade is close on time alone. It is decided by the other columns: rewriting is clearly worth it where you are applying through the company's own site to a role that fits, and clearly not worth it on a one-click posting you are unsure about.
Do applicant tracking systems reject most resumes automatically?
Not in the way the phrase suggests. Almost every large employer runs one, and what they mostly do is store, parse and rank rather than reject. The failure that costs interviews is parsing: contact details in a header, two-column layouts, and text inside images. If your resume copies cleanly into a plain text file, it will parse.
How long does a job search take?
The median was 108 days from first application to accepted offer, with about six days from application to first response and twelve from first interview to offer. Most of the time is spent between applications that went nowhere, not waiting on processes.
Why do people who send fewer applications get more interviews?
They do not. Searches that end at fifteen applications end there because the person was hired, and anybody still in the sample at a hundred is there because their rate was low. The statistic measures how long each search took and is frequently quoted as though it measured how well each person applied.
Should I cash out my 401(k) while I am out of work?
It is the most expensive money available to you. A withdrawal before 59½ is ordinary income plus a 10% penalty, so a third of it can be gone before it arrives. Leaving it where it is, or rolling it over, keeps the option open at no cost.