Do You Need a Different Resume for Every Job? We Compared 158,842 Pairs of Live Job Descriptions

“Tailor your resume for every job” is the most repeated advice in job hunting and the least evidenced. It is also expensive advice: taken seriously, it turns a hundred applications into a hundred rewrites. Anyone being asked to pay that cost is entitled to ask how different two job postings really are.

We can measure it. JobTether runs a scraper that reads postings straight from company career boards — the applicant tracking systems companies host their own openings on. On 29 August 2026 we took a snapshot of the entire pool and extracted the stated requirements from the text of 6,318 live job postings across 385 companies hiring in India. Then we compared postings against each other inside the same role: 158,842 cross-company pairs.

The five systems this pool is read from all publish their job-board interfaces: Greenhouse, Lever, Workable, Ashby and SmartRecruiters. Nothing here needs our scraper to reproduce.

The answer is not close. Two postings carrying the same job title at different companies share a median of 11.1% of their stated requirements. Widen to the same role and it falls to 7.1%. Out of 158,842 pairs, the number that asked for exactly the same things was zero.

What we measured, and what counts as a requirement

For each posting we read the requirements section only— everything from the first requirements or responsibilities heading to the first benefits, culture or equal-opportunity heading. That cut matters more than it sounds. A company’s “about us” paragraph and its perks list are full of capability words that are not requirements, and because that copy repeats on every posting a company writes, a careless reading turns one company’s boilerplate into a market-wide finding.

Inside that section we matched against a fixed list of 399 nameable capabilities — languages, frameworks, cloud platforms, tools, methods, finance and HR and operations skills — because the question is about things a resume can actually carry. Free phrase extraction would have scored two postings as different because their marketing prose differs, which is not the claim being tested. 5,607 postings (89.4%) had an identifiable requirements section, and 4,130 named at least four requirements inside it. Those are the postings analysed. A posting names a median of 8.

This biases every number below towards similarity, not away from it. A capability outside our list is invisible in both postings of a pair — and the terms most likely to be missing are the idiosyncratic ones, the in-house platform, the niche domain, which are exactly the ones that differ. Every overlap figure here is an upper bound. The real divergence between two job ads is wider than what follows.

First finding: the job title is barely a category

Before comparing requirements we had to group postings, and the grouping itself produced a result. Among the 4,130 analysed postings there are 3,456 distinct job titles. Of those, 3,125 — 90.4% — appear on exactly one posting in the entire pool.

Not similar titles. Identical strings, once each. The Indian job market does not advertise a few thousand standard roles; it advertises thousands of one-off ones, each named by whoever wrote the requisition. Before a single word of a job description is read, the shared vocabulary a generic resume depends on is already thin.

Same title, different job

The strictest version of the test: take only postings whose titles are character-for-character the same after normalisation, from at least five different companies, and compare every cross-company pair. Pairs from the same company are excluded throughout — two postings from one employer share a template, and counting those would measure boilerplate rather than the market.

Overlap is the share of the two postings’ combined requirements that both name. If one asks for six things and the other for eight, and they agree on two, that is 2 of 12 — 16.7%.

Across those 1,181 pairs the median overlap is 11.1%. 90.6% of pairs share less than a quarter of their requirements, 41.6% share less than a tenth, and not one pair shares more than half. Twelve postings all called “Software Engineer” ask for 87 different things between them and agree on 6.3%.

Look at the fourth column. Thirteen Product Manager postings name 64 distinct requirements. That is the number that makes the tailoring argument concrete: the title told you almost nothing about which 64 you were walking into.

The same picture at scale, across 25 roles

Identical titles are the strict test but a rare one, so we repeated it on roles. Each posting was assigned to a role family by the same matcher that powers our own job search, then we kept only families where at least 60% of the titles genuinely carry the role name. That gate matters: the matcher gates on a core word, so an ungated “Data Analyst” family fills up with compliance analysts and an accounts receivable analyst, and any overlap measured across that mixture would be measuring our bucketing, not the market. 25 families survived, covering 1,617 postings.

13 of the 25 families, by posting count. “One resume covers” is explained in the next section.

Across all 158,842 pairs the median overlap is 7.1%. 94.5% of pairs share under a quarter, 61.7% share under a tenth, and 0.1% share more than half. The pattern in the table is the important part: overlap falls as volume rises. The roles with the most openings — Software Engineer, Engineering Manager, Product Manager — are the ones whose postings agree least. Breadth and standardisation point in opposite directions, which is precisely the trap in applying widely with one document.

The one-resume test

Overlap between two ads is a fact about the market. The question a job seeker actually has is narrower: how much of any given posting would one good general resume answer?

So we built one, as generously as possible. For each role we took the most-demanded requirements across every posting in that role, as many of them as a typical single posting names, and treated that as the applicant’s resume. This is a better document than anyone could write in real life — it is assembled with hindsight over the entire role, including postings that had not been published when the resume was written.

That best-case generic resume matched a median of 37.8% of the requirements in an individual posting (45.0% when the titles are identical). At the bottom of the table, a general Software Engineer resume answers 26.9% of the average Software Engineer posting.

Read that number carefully, because it is not a claim about your skills. It says that roughly six in ten of the things a specific employer asked for are not addressed by a document written for the role in general — and in most cases the applicant genuinely has some of them. They are in an old project, a line that got cut for space, a tool named in a different vocabulary. The gap is not competence. It is which evidence made the page.

494 Software Engineer postings, no majority requirement

The largest clean family in the pool is Software Engineer: 494 live postings from 143 companies, naming 278 distinct requirements between them. If any role were going to have a stable core, it is this one. Here is how often each of the top requirements is actually named:

Not one requirement appears in a majority of postings.The most demanded skill in the largest engineering role in the country is named by fewer than half the ads asking for it. Across all 25 families, the median number of requirements shared by 80% of a role’s postings is zero.

There is no core to write to. The “standard Software Engineer resume” is a document addressed to a consensus that does not exist.

What applying broadly actually costs

Put the two halves together. A single posting names a median of 8 requirements. Twenty postings for one role name a median of 86.5 distinct requirements between them:

Twenty applications is a slow week for anyone job hunting seriously. One fixed document, carrying its 8 to 15 claims, is being tested against a target that moves across a hundred. That is the arithmetic behind the silence that follows a batch of applications, and it has nothing to do with the applicant being unqualified.

What this does not say

It does not say your resume should change its facts. Nothing in this data supports inventing a skill because a posting named it, and doing so is the fastest way to fail a technical screen you would otherwise have passed. What changes between applications is selection and language: which of your real projects appear, which are cut for space, and whether the capability you have appears in the words the employer used for it.

It also does not say every application deserves equal effort. If a posting shares 5% of its requirements with your background, tailoring will not rescue it, and the honest read of a low overlap is often that the job is not yours to win. The value of measuring the gap is partly in choosing where not to spend the afternoon.

Three limits worth stating plainly. Our sample is companies that publish on their own applicant tracking systems, which skews towards organised, tech-forward employers and away from postings that exist only on job boards. Our 399-capability list is thinner for creative and trade roles than for technical ones — the one family where a generic resume covered everything was Video Editor, and that says more about our vocabulary there than about video editing. And we read what postings say, which is not always what gets someone hired.

We ran one sensitivity check against the obvious objection — that the result is driven by soft, vague terms like “communication skills” that any posting might mention. Dropping all 16 of those terms and re-running on hard, checkable requirements only, the median overlap moved from 7.1% to 7.1%, and one-resume coverage from 37.8% to 39.0%. The finding does not depend on the soft terms.

So what do you actually do

The practical version of a 7% overlap is smaller than “rewrite everything”. Your history is fixed and your evidence is finite. What changes per application is the top third of the page and the vocabulary.

Read the posting’s requirements section — the part we measured — and list what it names. Anything on that list you genuinely have should be visible above the fold, in the posting’s own words rather than your synonym for it: the same capability shows up as “CI/CD” in one ad and “release automation” in the next, and a keyword screen does not know they are the same thing. Everything the posting did not ask for moves down or comes out. Our step-by-step guide to tailoring a resume to a job description walks through that edit properly.

That is fifteen minutes a posting done well. At twenty applications a week it is five hours, which is why almost nobody sustains it and why the advice keeps being given to people who cannot follow it. The honest options are to apply to fewer jobs with real care, or to let something else do the rewriting — which is the problem we built the agent to solve.

This is the fourth study off the same live pool. The others measured how long postings sit open, how rarely they state pay and how much experience they demand. Together they describe a market that asks applicants for precision it does not offer in return.