Greenhouse resume parsing, as Greenhouse documents it

Greenhouse publishes more about its parser than any other major vendor, including a named list of the formatting that breaks a parse. This page is that list, plus what sits around it.

The basics, quoted

"When you add a candidate to Greenhouse Recruiting, you can upload a resume to try and quickly fill in the information for their candidate profile. When you parse a resume, Greenhouse Recruiting scans an imported resume and auto-fills appropriate fields with information it detects when adding a new candidate or prospect to your account."

Greenhouse Support, Unsuccessful resume parse

Two things follow immediately. The parse runs on import, so it happens when the resume reaches Greenhouse, not when a recruiter searches. And it auto-fills fields, which means what matters is whether your text lands in the right field, not whether a human can read it.

The rules Greenhouse states

The full list of formatting Greenhouse names as a parse failure

This is the most useful thing Greenhouse publishes. Reproduced from the support article, with nothing added and nothing removed.

• A resume with spaces between the letters. "While it may appear cohesive to the naked eye, the parser won't recognize the separate letters as a single word and can't make sense of the data."
• Resumes that include graphics, photos, or word art.
• Resumes that are uploaded as an image, rather than a document such as a .docx or .pdf.
• Complex resumes with tables, headers, and footers.
• Resumes with the name and contact information in the header, footer, or text box.
• Resumes that have a columned layout.
• Resumes without clear sections and differing formats throughout each section.
• Company names that don't include identifying words such as Inc., Co., LTD, or LLC.
• Resumes with incomplete job titles, such as "Sr. Account Exec" instead of "Senior Account Executive".

Greenhouse is explicit that these "may result in a partial resume parse that will need to be manually corrected and verified." A partial parse is the outcome to plan for: some fields filled, some blank, and no error shown to the candidate.

The placeholder-data rule

Greenhouse also documents that it discards data it judges to be fake, and the list is specific enough to be worth reading before you anonymise a resume.

"The phone number and the first line of the address will parse correctly, since that information can't be definitively proven as real or fake. However, the name First Last or the Town name won't be recognized as real data and will be automatically skipped by the parser."

"Other incorrect data might include fake company names (such as Company 1, Client 1, Any Company), job titles (such as employee 1), school names, or other poorly designed data."

Greenhouse Support, Unsuccessful resume parse

Languages Greenhouse can parse

Greenhouse states that "full parsing capabilities are available" in the following languages, with support for regional differences such as Canadian French:

Arabic · Chinese (Simplified) · Croatian · Czech · Danish · Dutch · English · Finnish · French · German · Greek · Hebrew · Hungarian · Italian · Japanese · Korean · Norwegian · Polish · Portuguese · Romanian · Russian · Serbian · Simplified Chinese · Slovak · Spanish · Swedish · Ukrainian · Vietnamese

Source: Resume parsing with non-English languages. If your resume is in a language on this list, a language mismatch is not your problem. If it is not, expect a worse parse than the same layout would get in English.

What Greenhouse does not document

• No field-by-field mapping. Greenhouse lists which failure modes break a parse, but not the order in which it looks for a name, or what it does with a line it cannot classify.
• No keyword list and no ranking rule. There is no published set of terms that improve a match. The Talent Matching FAQ is about data minimisation, not matching logic.
• No accuracy figures. Greenhouse documents that parsing "can make mistakes" and that a parse can be partial, but publishes no rate.
• No per-employer configuration. Whether a particular company has parsing switched on for external candidates, or which fields are hidden, is not visible from outside.

The version of this that fits on a checklist

1. Export a text-based PDF. Keep it under 2.5 MB.
2. No photo, no icon next to your email address, no skill rating bars, no word art.
3. Your name and contact details are in the body of the document, in ordinary paragraphs. Not a header, not a footer, not a text box.
4. One column. No sidebar, no parallel skills column.
5. No tables used for layout. Section headings in the same place in every section.
6. Employers written out with their suffix where they have one: Northwind Labs Inc., not Northwind.
7. Titles spelled out: Senior Product Designer, not Sr. PD.

Everything above is a restatement of a rule Greenhouse published. The print-safe template in the gallery satisfies all seven, and the minimal template satisfies all seven but is harder to scan by eye.

• Greenhouse sources, all linked above Unsuccessful resume parse — the 2.5MB limit, the failure modes, the placeholder-data rule
  MyGreenhouse FAQ for Candidates — PDF guidance and the candidate review warning
  Resume parsing with non-English languages — the supported language list
  Talent Matching Data Processing FAQ — the fields Talent Matching extracts
  Manually add a candidate or prospect — where parsing is triggered
  Anonymize resumes — the supported way to blind a resume