What happens to your résumé before a recruiter sees it

Between your click on "apply" and a recruiter's eyes, your résumé passes through a pipeline: extraction → field mapping → profile creation → search index. Each step is mechanical, each can silently damage your application, and none of it is visible to you. This is a walkthrough of that pipeline from the parser's side — because we build one — with the failure mode of each step.

Step 1: extraction — your file becomes a stream of text

First, the system opens your file and pulls out raw text. For a .docx this is fairly safe: the document's structure — paragraphs, styled headings, order — is stored in the file itself. For a typical résumé PDF there is no stored reading order (tagged PDFs can carry one — most résumé exports don't); the parser reconstructs it from character coordinates on the page, which is educated guessing.

What can go wrong here: multi-column layouts interleave line by line; text inside images is invisible without OCR; icon fonts turn into garbage characters. The output of this step is typically a flat sequence of lines — and every later step builds on that sequence, right or wrong.

Step 2: field mapping — the text becomes a form

Next, the system tries to understand the stream: this line is a name, this token is an email, these lines are one job with these dates, this block is education. It leans on conventions — contact details near the top, standard section headings like "Experience" and "Education", date ranges next to titles.

What can go wrong here: your name is split across decorative lines and no single line reads as a full name; your contact details sat in a document header that extraction skipped; a creative section heading ("My Journey") gives the mapper nothing to anchor on, so the whole section lands in the wrong bucket — or in none.

Step 3: the profile — what the recruiter actually works with

The mapped fields become your candidate profile: a structured record with your employment entries, skills, and contact information. Here is the part worth internalizing: this reconstruction, not your designed PDF, is typically the working document of your application. It is what appears in lists, what gets compared side by side, what the search runs against.

The original file is usually attached and one click away — but a recruiter triaging a large pool works from profiles. If yours reconstructed with a missing phone number and a scrambled job history, that is the first impression, regardless of what the attachment looks like.

Step 4: search — where you surface, or don't

Finally, your profile joins the pool that recruiters query: skill terms, title terms, sometimes filters on years or location. Surfacing depends above all on two things this pipeline already decided — whether your text parsed into the right fields, and whether it contains the words being searched — alongside recruiter filters and anything you typed into the form directly. A perfect candidate with a broken profile, or with vocabulary that never overlaps the posting's, sits in the database unseen.

If this step is where your applications die, the diagnosis and fix are covered in the three failure modes.

What this means for how you write a résumé

• Write for the stream: one column, top to bottom — the order you want is the order stored.
• Feed the mapper its anchors: standard headings, name and contact as plain first lines of the body, dates next to titles.
• Speak the search's language: the posting's own terms, where they are true of you.
• Inspect the output, not the input: the version machines act on downstream is the extracted one — so look at it before an ATS does.