Getting Past the ATS: Keyword Matching for Indian Job Applications

There is a persistent belief that applicant tracking systems score your resume out of a hundred and reject anything under some threshold. That is not what happens, and believing it leads people to do genuinely counterproductive things — white text keyword blocks, twenty skills they cannot defend, "ATS-optimised" templates that parse worse than a plain document.

Here is what actually happens at most Indian companies. Your resume is parsed into structured fields — name, contact, employers, dates, education, a bag of skill terms. Those fields go into a database. A recruiter with a role to fill then queries that database, reads the first thirty or so results, and shortlists from them.

That is the whole mechanism. Everything useful about keywords follows from it.

The query is shorter than you think

Ask a recruiter what they typed to fill their last role and the answer is rarely elaborate. It is usually two to four terms and a filter:

"Java" AND "Spring Boot" AND ("Bangalore" OR "Bengaluru"), experience 3–6 years

Or, on Naukri or LinkedIn Recruiter, a job title plus one non-negotiable skill plus a location. The reason it is short is the same reason your resume gets twenty seconds: the recruiter has forty roles open and no interest in constructing an elegant boolean. They add terms only when the result set is too big, and drop them when it is too small.

Two consequences, and they point in opposite directions from the advice most people follow:

1. A handful of terms decide almost everything. Not fifty. The three or four skills that define the role, the title, and the location. If those are missing or written differently from how the recruiter types them, nothing else on your resume gets a chance.
2. Extra keywords do not help. They do not raise a score, because there is no score. What a long undifferentiated skills list does is make you appear in searches for things you cannot do, where you will be screened out by a human in the first five minutes — which costs you nothing on that role, but it does teach the recruiter that your resume is unreliable.

Read the job description as a query, not as prose

The posting is written by someone who will also be searching. The vocabulary in it is very close to the vocabulary in their search box. So work backwards from it.

Take the posting and mark, in this order:

• The title. If it says "Data Analyst" and your last title was "MIS Executive", that gap is real and you have to close it — more on that below.
• Anything under "must have" or "required". These are the AND terms.
• Named tools and technologies. Power BI, SAP FICO, Figma, Kubernetes, Tally, Salesforce. Proper nouns are what people search for, because they are unambiguous.
• Certifications and qualifications stated as hard filters. CA, CFA Level 2, PMP, B.Tech in a named branch.

Then check each one against your resume, literally. Not "do I have this skill" — "does this exact string appear on my document".

The exact-match problem, and where it bites in India

Search is largely literal. A parser does not know that these are the same thing:

Some systems handle some of these. Many handle none of them. The fix costs nothing: write the long form and the common short form once each, in a place where it reads naturally.

B.Tech (Bachelor of Technology), Computer Science — VTU, 2021

Skills: Power BI, SQL, Advanced Excel (Power Query, pivot models), Python (pandas)

Bengaluru, Karnataka (open to Bangalore-based hybrid roles)

That last one is slightly awkward and I still recommend it, because the Bengaluru/Bangalore split genuinely fragments search results, and recruiters at older companies type the old name.

Where a keyword sits changes what it is worth

The same word carries different weight depending on where it appears, for a reason that has nothing to do with algorithms: it is where the human reader looks.

In your title line. The strongest position available, and the most underused. If you are an "MIS Executive" applying for analyst roles, do not invent a title you never held — put the market-standard name next to the real one:

MIS Executive (Data Analyst) — Ashirvad Pipes, Bengaluru | Jun 2023 – Present

This is honest, it is common practice, and it puts you in the result set for "data analyst" where your real title would have hidden you.

In your bullets, attached to work. This is where a keyword becomes credible instead of merely present. Compare a skills line reading "SQL" against:

Rewrote the weekly channel-performance report as a set of SQL views on Redshift; cut the manual build from four hours to a scheduled refresh, now used by three regional sales heads.

Both make you appear in a search for SQL. Only one survives the recruiter reading it.

In the skills section. Necessary, useful, and the weakest of the three on its own. Treat it as the index, not the argument.

Nowhere near the header or footer of the page. Many parsers skip document headers and footers entirely. People put their phone number there and then wonder why nobody calls.

Tailoring, in ten minutes per application

Rewriting a resume per application is unsustainable and nobody does it for long. What works is a small, bounded edit that takes ten minutes:

1. Keep a master resume — everything you have ever done, four pages, never sent to anyone.
2. For each application, copy it and cut to one page, keeping the bullets closest to the posting.
3. Rewrite the summary line to name the role you are applying for.
4. Reorder the skills line so the posting's must-haves come first.
5. Adjust two or three bullets to use the posting's vocabulary for work you genuinely did.

Do this for roles you actually want. For everything else, send the general version — the marginal return on tailoring the fortieth application does not justify the hour.

One thing worth doing every time, because it is free: name the file properly. Rakshith-Gowda-Data-Analyst.pdf beats resume_final_v3(2).pdf in a recruiter's downloads folder, and some systems index the filename.

The things that break parsing

Before worrying about keywords at all, make sure the document parses. In rough order of how often I see each one cost somebody an interview:

• Two-column layouts. The single most common self-inflicted wound. A designer template with a sidebar of skills often parses into interleaved fragments — half a job title, then a skill, then a date. Use one column.
• Skills as a graphic. Star ratings, percentage bars, tag clouds rendered as images. To a parser these are nothing at all.
• Tables holding your work history. Some parsers handle tables; enough do not.
• Scanned or image PDFs. If you cannot select the text in a PDF reader, neither can the system. This includes a resume you signed and scanned.
• Unusual section headings. "My Journey", "What I Bring". Use Work Experience, Education, Skills, Projects.
• Dates in inconsistent formats. Pick Mon YYYY – Mon YYYY and use it everywhere, including "Present".

Test it yourself in thirty seconds: open your PDF, select all, copy, paste into a plain text editor. What you see is roughly what the parser sees. If it is scrambled, no amount of keyword work will help.

What not to do

Do not paste the job description into white text at the bottom of the page. It is detected, it is treated as deception, and at companies with any recruiting maturity it gets you blocked rather than merely rejected. The gain if it worked would be one appearance in one search; the cost is your name on a list.

Do not list skills you cannot discuss for five minutes. Every term on your resume is an invitation, and technical interviewers accept the ones you look least confident about. A shorter, defensible list outperforms a long one, consistently.

Do not chase a keyword you do not have. If a posting requires three years of Kubernetes and you have read about it, you are not going to argue your way past the person who does have it. Spend that hour on a posting where you are a real candidate.

The portal versions of this

Naukri is the biggest ATS in India in practice, because for many roles the recruiter never leaves it. Your profile there is your searchable record, not your uploaded PDF, so the keyword work has to be repeated in the profile fields — key skills, IT skills, current designation, preferred location. The resume headline is indexed and is the closest thing to a search-visible title you control. Profiles that have been updated recently also surface higher in many recruiter views, which is the only real reason to log in periodically.

LinkedIn ranks on the headline, the About section and the skills list, in roughly that order. The headline is 220 characters and most people waste them on "Aspiring Data Scientist | Passionate Learner". Put the role and the tools in it instead. If you want the longer version of this, it is in LinkedIn for the Indian job market.

The honest limit of all of this

Keyword matching gets your resume in front of a person. That is the entire benefit, and it is worth having — a resume nobody sees loses to a mediocre one that gets read. But it is a distribution problem, not a substitute for the document being good, and the twenty seconds after the search is what actually decides the shortlist. The format, the bullets and the cuts that matter are in the resume guide.

If you are applying at the start of your career, the same mechanics apply with one difference: your keywords come from projects rather than job titles, which is a weaker signal in search and a stronger one in the interview. There are fresher roles listed here, and the specifics of that resume are in the fresher resume guide.