Resume and calls

An interview starts with a call, and the call starts with a chain of readers. Software parses your file into fields, a recruiter searches those fields and your profile by keyword and filter, a person skims the top third for seconds, and an interviewer later reads every bullet as a question. A call then comes through one of three doors: a search that finds you, a person inside the company who submits you, or a short message to a stranger.

The file: parsed, searched, skimmed

For: everyone.

No vendor document read for this lesson (Greenhouse, Workday, Lever, Oracle Taleo) describes a hidden "ATS score" that rejects a resume. They describe a parser that fills fields from your file, and recruiters who search and review those fields. A knockout question on a form (notice, location, permission to work) can reject automatically; a resume score is unverified. A fault at one reader cannot be repaired at the next: a fine bullet does nothing if the parser put your employer in the title field and the search never found you.

Across vendors, an image-only file gives the parser nothing and the output is fields, possibly partial or wrong; only Greenhouse names the layout risks.

Check your file. Select all in the PDF, paste into a plain-text editor and read top to bottom: your name first, every heading in order, each date beside its role. If nothing selects, it is a picture. Use a text-based PDF (or DOCX if the company asks), under the stated upload limit (Greenhouse 2.5 MB; Naukri's FAQ asks for under 2 MB, check the upload screen). The paste test is the author's reading, not a vendor method.

Names, titles, dates. The same three facts are read by the parser now and by a background-verification vendor at the offer, who compares them with your letters and your provident fund record. Write the employer's legal-style name once as your letter has it, your letter's designation first with your working title in brackets, and month and year on every role. A gap is one line with dates and a fact ("Career break, Mar-Oct 2023: family caregiving"); client work under an NDA is your employer's name plus the client by sector. Leave out photo, date of birth, marital status and a "declaration"; give any of them in a form that asks. All of this is the author's reading.

Next stop: What a recruiter searches, 2 min

What a recruiter searches

For: everyone.

Beyond applications, recruiters search Naukri's recruiter database (Resdex), LinkedIn Recruiter and their own system. A search is keywords plus filters and returns a list; if you are not on it, nobody skims you. Resdex offers any, all, excluding and mandatory keyword modes, and filters for location, experience, designation, company, salary and notice period. LinkedIn Recruiter filters on job titles, location, companies, skills and more, marks each term Can have, Must have or Doesn't have, and takes Boolean in the keywords field; its help pages show no notice-period filter. Naukri's job-seeker FAQ says recruiters tend to favour recently updated profiles.

Most filters read profile fields; the resume is searched only as keywords. So the resume must carry the keywords as plain words, and agree with the profile on titles, companies and dates, because a person who opens both notices a difference. Type notice and CTC fields from your documents, the same on every site; the rules are in lesson 2. A field typed to pass a filter becomes the first thing the call checks.

Check. Write the five words a recruiter for your target role would type, then find each as a plain word in your profile and your resume. On Naukri, Profile Performance shows the terms that found you. On LinkedIn, set Open to Work to recruiters only, mark your current employer correctly, and write nothing you would not say in your notice meeting; LinkedIn says it cannot promise complete privacy from your employer.

Next stop: The skim, 1 min

The skim

The "seven seconds" comes from a US site, Ladders: HR Dive reported on 8 November 2018 an average of 7.4 seconds on the initial screen in an eye-tracking study. No sample size was reported, the study was not opened, and no Indian equivalent was found. Read it as "short". Clear sections, bold titles and bulleted accomplishments did well; clutter, columns and long sentences did poorly. Amazon's recruiters, in a company article, say: actions and results, accomplishments rather than responsibilities, numbers where you can, keywords from the posting, no decoration.

Length is the author's reading: one page to about six years. From six to eight, ask one question: would a second page hold new evidence of scope (a system other teams depend on, people you hired), or only more bullets of the kind page one proves? From about eight years, two pages, with older roles in two lines each. The top third carries level, domain, scope and one number (the first rows of the table in "One resume, before and after"). Test it: hand the page to a friend for ten seconds and ask level, domain, one result.

Next stop: Bullet grammar, 2 min

Bullet grammar

Five slots, in priority order: verb (Moved, Cut, Built, Led), object (the system), change (a measured before and after, a rate or a count), method (how, in one clause) and scope and ownership (who used it, how big, your part). The rules, all the author's reading:

• One idea, one or two lines, 20 to 30 words. Lead with the result or the verb.
• Every number defensible: measured how, over what window, from what source. Never round up. With no number, use a count or a before and after in words.
• Cut "worked on", "responsible for", "helped", "involved in", "significantly". Tools go in the skills line.
• A team result stays a team result: "with two engineers", "led", never solo credit.

Bullet check. A bullet should pass at least five of six: a specific past-tense verb; a before and after, count or named outcome; one clause of method; who or how many and your part; "measured how?" answered; two lines or fewer. The cut-off is the author's estimate (not measured). Numbers an AI made up are the commonest fault: if you did not measure it, delete it.

Next stop: One resume, before and after, 2 min

One resume, before and after

Kavya (six years, SDE-2) and Meera Subramanian (ten years, SDE-3) are invented, with every employer and number. The full pages are in the appendix.

For: SDE-3, Staff, Principal.

Three differences decide the level, the author's reading. A senior bullet names a decision and what was rejected; the mid one lists work. It shows who else moved because of you. And it reports a number a year later, not on launch day. Name a service-firm role by its module, size and a before and after, with the client by sector. Two moves inside one employer go on one entry with both dates, so no one reads a hop. Choose two referees who saw the decision and the result, ask them first, and send them the three bullets you most want checked. Use title words you can defend ("designed", "tech lead"): a recruiter filtering for "Lead Engineer" cannot find "Senior Software Engineer II" unless the lead word is also in your profile.

Tailor by reordering, not rewriting. Keep a base file with every bullet. For each posting, mark each qualification true, partly true or not true of you, move the three closest bullets up, and use the posting's words only where they are true ("Java, not Kotlin yet" is a complete answer). Titles, employers, dates and numbers never change. Then check that your one-minute pitch and your top third say the same three things in the same order: what you own, one result with a number, why this next step.

Next stop: Where hires come from, 1 min

Where hires come from

Ashby studied over 38 million applications and, separately, nearly 250,000 hires (2021 to June 2025); Gem's 2026 report pools over 165 million applications; LinkedIn published referral figures in 2018. All are global and US-weighted, mostly technology firms; no named Indian survey was found, and vendors define stages differently, so figures from one cannot be multiplied by another's.

Read the direction, not the size. A referral gets you to a screen more often; it does not skip the loop: of 1,000 referred applicants, 400 interview and about 64 reach an offer stage (6.4%). When anyone gives you a funnel figure, ask whose data, which years, which country, and what is counted at each stage.

Next stop: Referral schemes and the 60-second ask, 3 min

Referral schemes and the 60-second ask

Few Indian firms publish their schemes. People Matters reported on 2 March 2026 that TCS relaunched "Bring Your Buddy", paying up to Rs 40,000 for mid-level hires (4 to 15 years), falling with the days from the candidate accepting to joining: Rs 40,000 for 1 to 30 days, 30,000 for 31 to 45, 20,000 for 46 to 60, 15,000 for 61 to 90. Trade press reported up to Rs 50,000 at Infosys (October 2025). Nothing official was found for other firms. The shape below is the author's reading; put each point to your referrer as "typically; check your policy".

1. The employee submits you on an internal portal against a job ID, usually with a short note on why you fit.
2. The reward is paid after you join, often after a minimum-service period, and falls with how soon you join.
3. If an agency has already submitted you, the employee's referral for the same role may be rejected as a duplicate. Keep a tracker.

Ask the four questions in Figure 3.5 before they submit. For a company where you know nobody, ask a connection for an introduction to an engineer, not a referral to the company: "Would you introduce me to Kiran by email, saying we worked together on refunds?" Then ask Kiran for fifteen minutes about the team. Never pay for a referral, and no one can promise an interview.

Cold messages (principles are the author's reading; SHRM's 2021 analysis of recruiters' InMails found 200 to 400 characters did best, but no dataset measures candidates writing to engineers). Name the job ID; give two facts you own, one with a number; make one small ask (two questions for a stranger, a referral only from someone who has seen you work); give an easy no, once; state your notice period if you ask for a referral; stay inside the limit (200 characters for a LinkedIn connection note); follow up once after five working days, then stop; write nothing confidential. Copy the structure, not the sentences.

Drill: the 60-second referral ask. Pick a real role and a real former colleague. Record yourself on your phone with a one-minute timer: who you are in one sentence; the role and its ID; two facts of fit; your notice and what is unresolved; the one thing you need; the easy no. Three takes; score the last, then write it as a message, count the characters and send the real one the same day. Score only what you can point to.

Run it as a pipeline. Write 25 to 40 target companies (the author's judgement) in the tiers from lesson 1, with a person you know, or one step away, beside each; a blank means portal-only for now. Start with practice companies. Keep one tracker: company, role and job ID, route, date, follow-up date, status. After three weeks it shows which door produces calls; in an invented tracker, 12 portal applications gave no calls, while 5 referral asks gave 2. Move your hours to the door that works.

Next stop: Pitfalls, 1 min

Pitfalls

• Rounded dates, thirty skills, borrowed bullets. Each fails a month-by-month check, a half-hour of questions, or "tell me more about that".
• Sixty portal applications in a weekend. Twenty targeted approaches, half referrals, tell you more by week three.

Practice with an AI: the prompt kit in the appendix has 3 cards for this lesson.

Sources