How Lever actually parses your resume (and why "orphan text" is its

Lever is the third-largest enterprise ATS by deployed seats and is especially common at consumer-facing companies and mid-stage SaaS. If you've applied to roles at companies like Netflix, Twitch, Quora, or Clover Health in recent years, your resume has been parsed by Lever.

Lever's parser is similar to Greenhouse's — both use third-party text extraction layers and prioritize structured candidate-profile population — but Lever is significantly stricter about "orphan text" (content not under a recognized section header). Resumes that survive Workday and Greenhouse sometimes lose half their content in Lever.

This post covers what Lever actually does with your PDF, the orphan-text problem in detail, and the formatting choices that survive Lever specifically.

How Lever differs from Workday and Greenhouse

If you've read How Workday parses your resume and How Greenhouse parses your resume, the high-level pipeline is similar — extract → structure → match. But Lever has its own tendencies:

Behavior	Workday	Greenhouse	Lever
Multi-column layouts	Catastrophic	Imperfect	Better, but image-only regions in side columns get skipped
Tables	Flattens to noise	Mostly handled, may lose labels	Handled, but cell content can split across roles
Section anchoring	Loose	Strict	Strictest — orphan text typically dropped entirely
Date parsing	Strict	Moderate	Lenient — handles more variations
Image-only regions	Skipped	Skipped	Skipped silently without warning
Candidate UI	Search-heavy	Pipeline board	Pipeline board with strong recruiter notes

The biggest practical difference: Lever drops "orphan text" — anything not clearly under a recognized section header — more aggressively than the other two.

What "orphan text" means in practice

Orphan text is content that exists in your PDF but doesn't sit under one of Lever's recognized section anchors. Examples:

A "Highlights" or "Featured Work" callout box at the top of the resume (between the contact header and Experience section)
Sidebar content (a left-column "About Me" or "Languages" panel)
Inline quotes or testimonials between roles
Project descriptions outside a "Projects" section
"Key Achievements" boxes that don't sit under "Experience"
Anything in the footer of every page (like a tagline or quote)

In Workday, orphan text often gets parsed as freeform "additional content." In Greenhouse, it goes into a "Notes" or "Other" field. In Lever, it most commonly disappears entirely from the parsed candidate profile.

The visible symptom: the candidate's Lever profile is missing major content the resume clearly contains. The recruiter has to open the PDF manually to see it (which they often won't do during initial screening).

The Lever parsing pipeline

Lever uses Sovren-style text extraction plus internal Lever logic. The flow:

1. Text extraction. PDF → plain text via standard parser. Reading order is left-to-right top-to-bottom (so multi-column layouts still produce interleaved nonsense, just like Workday).

2. Section anchoring. Lever looks for standard section headers and parses content under each. Recognized headers: Experience, Work Experience, Professional Experience, Employment History, Education, Skills, Summary, Profile, Projects, Certifications.

3. Orphan-text drop. Content that isn't clearly under a recognized header is silently dropped from the structured profile. The PDF is preserved as a downloadable attachment, but the recruiter-facing data fields don't include it.

4. Entity extraction. Within Experience: company, title, dates, location, bullet text. Within Education: institution, degree, date, location. These populate the candidate profile cards.

5. Pipeline scoring. When the role has tagged required skills, Lever scores keyword overlap. The score informs candidate ranking but doesn't auto-reject.

6. Recruiter notes integration. Lever's notable feature is structured recruiter notes — every interviewer's feedback is searchable. This means the resume has to land hard enough that the candidate even GETS to interview; once interviewed, the resume's role diminishes vs notes.

What Lever weights heaviest

Across mid-stage SaaS and consumer roles:

Tag-required skill keyword presence. Same logic as Greenhouse. Resumes that mention every required skill outrank ones missing any.
Recent role recency. Lever weights bullets from your current and most-recent roles much higher than older ones.
Title progression coherence. Lever flags title mismatches (e.g., "Senior" then "Junior" in next role) for recruiter attention. Also flags very long tenures at junior titles.
Section completeness. All four standard sections present is a positive signal.
Bullet format consistency. Lever's parser performs better on bullets that all use the same prefix style (all "•" or all "-" — not a mix).

The three things that break Lever parsing

After analyzing several hundred resumes through Lever-style parsing, three patterns dominate the failure modes.

1. Sidebar layouts (the biggest single Lever failure)

Sidebar layouts are popular in modern resume templates (Canva, Beautiful.ai, Notion templates). Typically a colored or shaded narrow column on the left or right contains: photo, contact info, skills as chips, languages, hobbies, or "About Me" text.

Lever's parser reads the document top-to-bottom in linear order. The sidebar content gets interleaved with the main column, then most of it gets classified as orphan text and dropped. What survives: usually the main-column Experience section, partially. What gets dropped: the entire sidebar content, plus any main-column content that the orphan-text classifier confuses for sidebar.

The visible symptom: the candidate's Lever profile has Experience but no Skills section, no contact info beyond what's in the main body, and no Education if Education was in the sidebar.

The fix. Single-column layout. If you want a visually distinguished section (like Skills as chips), keep it in the main column with a clear "Skills" header above it, formatted as plain text.

2. Image-only regions in the resume

Lever skips image-only regions silently. Common cases:

A photo with text overlay (your name and title rendered as part of the photo image)
A skills "infographic" with bars or radial charts
A QR code linking to your portfolio
A logo for each previous employer rendered as image
A "language proficiency" widget with flag icons

The visible symptom: the recruiter sees a much sparser parsed profile than what the PDF actually contains.

The fix. Anything important should be plain text. Photo is optional (depends on region; in US/UK, omit). Skills are plain text comma-separated lists, not visualizations. Portfolio link is a plain URL in the header (Lever extracts URLs from text, but not from QR codes).

3. Creative section names (same as Greenhouse, more punishing)

Lever uses essentially the same section-anchor regex as Greenhouse. The same creative-naming problem applies — but Lever's silent-drop behavior makes it worse: in Greenhouse you might at least get content into a "Notes" field; in Lever it just vanishes.

The visible symptom: candidate profile shows "Experience: (none extracted)" even though the resume clearly has work history.

The fix. Use the standard recognized headers. "Experience" / "Education" / "Skills" / "Summary" all work. Decorative tag-lines can sit BENEATH the standard header:

Experience
Where I've made impact

[role bullets...]

How to test your resume against Lever

The free LSI Resume Analyzer runs a Lever-style parser simulation alongside Workday, Greenhouse, Taleo and plain-text. You'll see per-engine cleanliness scores and the specific structural failures (sidebar layouts, image-only regions, orphan text, creative section names) that would happen in Lever on your file. Runs entirely in your browser.

If you want to test against an actual Lever instance:

Find a job at a Lever-using company (their job board URL typically contains jobs.lever.co/companyname).
Apply with your resume.
Lever often has an "auto-fill from resume" feature on the application form. Watch what fields populate. Missing fields = parsing failures.

Most Lever applications don't show you the parsed profile, but the auto-fill behavior is a reliable proxy.

Quick checklist for Lever survival

Single-column layout (no sidebars)
Standard recognized section headers ("Experience" / "Education" / "Skills" / "Summary")
Plain-text content (no image-only regions, no QR codes, no infographics)
All bullets use the same glyph (• or -, not mixed)
Real text PDF (verify by copy-pasting from your PDF in any reader)
Photo only if culturally expected for your geography (omit for US/UK)
Standard date format ("Month Year — Present" or "MM/YYYY — MM/YYYY")

When orphan text is actually fine

A nuance worth flagging: not all orphan text is bad. Some content can sit between sections without causing problems:

A 1-line tagline directly under your name in the contact header (Lever extracts the contact area as a unit; brief content here is fine)
The Summary section content (if you have a "Summary" header)
A horizontal divider line (purely visual, no text content)

The problem is blocks of substantive content that aren't under a recognized header. Single lines or brief decorative text usually survive; paragraphs of orphan text usually don't.

For broader ATS context, see How an ATS Reads Your Resume. For role-specific keyword libraries Lever will check against, see /resume-keywords.