11 min read
You didn't rewrite your resume between Tuesday and Thursday. Company A sent a screen invite. Company B went silent. Company C auto-rejected before lunch. That's when ATS decisions feel random, and it's brutal when you're already tired from applying.
Here's what actually changed: not your file, but the req filters, posting age, and parser state on each employer's side. Greenhouse on one team may hard-filter on title plus years. Workday on another may import your skills into the wrong column and drop you before keyword math even runs. Same PDF. Three different rule sets. None of them are rolling dice on your career.
Before you burn a weekend rewriting bullets that already work, check your resume for free against the posting and confirm what the portal preview imported. If title and dates parse clean on one site and break on another, you're looking at parser state, not bad luck.
This page walks five steps to audit those three variables, two edge cases where identical files diverge inside one company, where apply logs still lie to you, and a copy-paste tracker you can run in ten minutes tonight.
Quick Wins
- Log ATS name, req age, and preview screenshot on every submit.
- Paste your export into Notepad before you blame keyword gaps.
- Tailor bullet one to each posting's must-haves, not a master file.
- Track pass, fail, and silence in one apply log row per company.
What changed when ATS decisions feel random
Random is the story your brain writes when inputs hide. Applicant tracking doesn't flip a coin on your PDF. Each req carries saved filters a recruiter or coordinator set before the posting went live: title strings, years in role, license keywords, location radius, and sometimes a cap on how many profiles move to human review.
Three levers move between employers while your file stays still. Req filters decide whether you enter the list. Posting age decides whether early files get skimmed before volume spikes. Parser state decides whether the system even sees the words you typed in Experience or dumps them into a certificate line.
A composite mid-level data analyst uploads the same single-column DOCX to two corporate portals on the same afternoon. Greenhouse imports SQL in Skills and passes a loose keyword bar. Workday attaches the analyst title to a certificate block and fails a seniority filter set to five-plus years. Same qualifications. Different imported text. Different outcome. That pattern is predictable once you log the variables.
Posting age adds another layer. Some teams review in arrival order for the first forty-eight hours. Others only open the req after a hiring meeting on Friday. A file that feels late on a flooded req may still win on a fresh posting where filters are wide and volume is low.
Parser state is the lever candidates skip. Two-column templates, text boxes, and icon bullets scramble import order. You can pass a keyword checker on the PDF while the portal preview shows blank title rows. Recruiters search parsed fields, not your attachment, on the first pass.
Score widgets on third-party sites add noise here. They rank your PDF in a vacuum. Corporate reqs rank imported text against filters someone saved last month. When your checker says strong match and the portal preview shows a blank employer row, trust the preview.
Read how hiring software supports recruiters when you need the map of which fields get searched before anyone opens your PDF.
Why ATS decisions shift: five-step audit on one resume
Run these steps in order on your last three applies. You're hunting rule differences, not rewriting your whole career story.
Step 1: Log the three variables before you submit
Open a notes row per apply. Write employer name, ATS if listed on the careers page, date the req posted, date you applied, and whether the portal showed a preview after upload. Those four facts explain most swings candidates call random.
Before: Same master PDF sent to twelve reqs with no notes. Silence on eight feels like a broken file.
After: Row reads Greenhouse, req age two days, preview clean, pass to human review on one apply. Row reads Workday, req age nineteen days, title field blank in preview, auto-filter on seniority for another.
Patterns show up fast when you stop treating each silence as a unique tragedy. Req age and parser state belong in the same sentence as keyword match.
Step 2: Compare portal preview to your export
Upload once. Screenshot the parsed view: title, employer, dates, top three bullets, Skills. Paste the same DOCX into Notepad on your desktop. Side by side, mark any line that merged, vanished, or landed under the wrong heading.
Before: Registered nurse resume with title trapped in a sidebar. Portal preview shows blank Title and imports license under Skills.
After: Single-column layout with Registered Nurse, Med-Surg above employer name. Preview shows title on its own row; bullet one reads Documented care in Epic on 32-bed unit; precepted four orientees on fall-prevention workflows.
Parser state beats keyword density when import drops your proof. Fix layout before you add a fourth synonym for stakeholder management.
Read why your resume looks different after Workday upload when dates and titles flatten for the same text-box reasons on US corporate portals.
Step 3: Read posting age and queue signals
Check when the req went live and whether the board still says accepting applications. Fresh reqs with low volume often get human skims in posting order. Old reqs with hundreds of imports may never reopen unless a filter narrows the list.
Note reposts. Some teams close and reopen the same role with tighter filters after a bad first batch. Your file did not change. The saved search did. I've seen coordinators add a license string on day ten after too many unqualified imports. Files that passed on day two would fail the same req on day eleven without a single word edited on your side.
Early apply helps when humans still scroll arrival order. It does not override hard filters. Uploading first on a req capped at Senior when your parsed title reads blank still ends the same way.
Step 4: Mirror req filter language in bullet one for that employer
Open the posting. Highlight three must-have terms from the first three bullets in the job description. Rewrite bullet one under your current role so one term lands in the first eight words. Keep one master layout. Swap language per req, not per fantasy version of your career. Save each tailored export with the company name in the filename so you never re-upload last week's version by mistake.
Before: Operations coordinator applying to a logistics req with bullet one about generic process improvement and no warehouse system named.
After: Bullet one reads Coordinated inbound scheduling in Manhattan WMS for 12 dock doors; cut late trailers 14% in Q2. Skills repeats WMS as backup.
Filters often search title plus one tool plus years. Your headline should carry the title noun. Bullet one should carry the tool. Dates should show tenure for the level line. Missing any leg drops you on that req even when another req would have passed the same PDF yesterday.
Step 5: Track outcomes in a simple apply log
After each submit, write one row: company, ATS, req age, preview status, filter guess, outcome in seven days. Outcomes are pass to human, auto message, or silence. Silence after a clean preview on a young req points to competition, not parser failure. Silence after a broken preview points to layout.
Copy-paste apply log header you can drop into Notes or a spreadsheet:
Company | ATS | Req posted | Applied | Preview OK? | Filter notes | Outcome (7d)
Example: Northwind | Greenhouse | Sep 2 | Sep 4 | Yes | Title+SQL+3yr | Screen
Example: Acme | Workday | Aug 10 | Sep 4 | No title row | Senior cap | Silence
Read why your resume scores high but you still get rejected when parsing is clean but human skim or filter caps still block the next step.
Edge case: two reqs at one company, different filter sets
Large employers run multiple ATS instances or req templates by division. Your operations file may pass a corporate Greenhouse req with loose years filters and fail a subsidiary Workday req with a hard license string the parser hid under Certifications. Log division name, not just company name, on each row.
Edge case: re-upload after a layout fix on the same req
Some portals overwrite parsed fields cleanly on a second upload. Others create duplicate candidate rows that confuse coordinators. Screenshot the first preview. Fix layout offline. Prefer a fresh apply on a reposted req when the first import scrambled title or dates. Do not assume the new score retroactively fixes the first row recruiters already skimmed.
Where randomness stories break down
Blaming the ATS instead of the req filter. Auto-reject messages often mean seniority, location, or license gates, not a mystical score.
Using one master file for every posting. Filters differ. Bullet one should swap terms per req while layout stays plain.
Ignoring preview after upload. Keyword-rich PDFs fail when title imports blank. Fix parser state first.
Assuming early apply overrides bad layout. Arrival order only matters after import succeeds.
Chasing score widgets without logging outcomes. A checker helps structure. Your apply log proves which lever moved.
Re-uploading five times on one broken row. Duplicate profiles annoy coordinators and rarely repair scrambled dates.
Treating silence on old reqs like silence on new ones. Volume and closed pipelines behave differently. Req age belongs in every post-mortem.
Confirm parser state before the next batch
Structure first, then tailor bullet one per req. A parse check catches layout breaks that make filters look random when import dropped your proof.
Run a free ATS resume check with the job description pasted in. Fix title rows, date lines, and table flags before you log the next apply row.
Use job match score to decide whether a req deserves a bullet-one tailor pass or a skip when filters clearly outrank your tenure line.
Stop treating ATS decisions like lottery tickets
ATS decisions feel random when you can't see req filters, posting age, or parser state moving under the same PDF. Log those three levers on every apply and the pattern stops being luck. It becomes a checklist you can act on tonight.
Pick your last silent apply. Pull the preview screenshot if you have it. Paste your export into Notepad. Write the req age and filter guess on one log row. Fix layout if title or dates broke. Swap bullet one if import was clean but the posting asked for a tool you buried in bullet four.
Job searching is already exhausting. You don't owe anyone a perfect file on every portal. You do owe yourself evidence before you rewrite bullets that already proved scope on a different ATS import. One hour logging beats one weekend guessing.
Batch your applies by parser family when you can. Run the Notepad paste once, then swap bullet one language across three Greenhouse reqs before you touch a Workday export.
This won't turn a reach req into a hire when years filters cap below your tenure. It does stop qualified files from dying because Workday read your title as blank while Greenhouse on another tab passed the same words an hour earlier.
Keep the apply log for two weeks. When three silences share broken previews, fix layout once and re-run the batch. When three silences share clean previews on old reqs, shift energy to fresher postings instead of another synonym pass.
Save one folder per target company with preview screenshots beside plain exports. Side-by-side files settle arguments about whether the parser or the filter moved.
And when a recruiter replies, ask which filter knocked similar files out this week. Their answer calibrates your log faster than any generic ATS myth thread. Most candidates never ask because silence feels personal. Treat it as data until proof says otherwise.
When a portal asks for a short note after a confusing auto message, generate a cover letter that names the filter you cleared in bullet one so a human sees context the parser skipped.
- Log ATS, req age, and preview status on every row.
- Fix parser layout before keyword rewrites.
- Tailor bullet one per posting while master layout stays plain.
Run the five-step audit on your last three applies, then run the free resume check on the next req before submit. Same resume, different outcomes. Now you'll know which lever moved.
Read more
Frequently asked questions
No. Each employer runs different req filters, posting windows, and parser imports on the same PDF. Greenhouse on one req may require a title string Workday on another ignores. Your file did not change. The rules attached to that application did.
Keywords are only one filter leg. Posting age, seniority caps, location rules, and how the parser read your layout on upload also gate the list. A mid-level analyst file can pass a Greenhouse req with loose years filters and fail a Workday req set to Senior only before a human opens either PDF.
Early helps when recruiters review in arrival order or close the req at a volume cap. It does not override hard filters. A file uploaded on day one still drops if the saved search requires a license string your parser placed under Skills instead of Experience.
Fix layout first, then submit once on a fresh export. Some portals overwrite parsed fields on re-upload and create duplicate profiles. Screenshot the preview after the first upload. If title or dates import wrong, flatten the DOCX, confirm the Notepad paste, and apply to a new req rather than fighting a broken import twice on the same row.