Does a two-column resume break parsing? We measured it

Can an ATS read a two-column resume?

Sometimes, and you cannot tell which time yours will be. We laid out the same 36 resumes in one column and in two, and read every PDF back the two ways parsers read a page: in the order the file stores its text, and line by line across the page. One column came through both readers intact — all 372 fields and all 174 bullets. With a sidebar, each reader broke something different, and which one an employer's software uses is not something it tells you.

The results

The fields are the ones a recruiter needs to reach you and read your history: name, email, phone, and each role's title, employer and dates, graded by the same scorecard the builder's X-Ray tab uses. A bullet counts as in one piece when its whole sentence appears unbroken in the extracted text. The words, the font and the page size were identical across the three layouts; only the arrangement changed.

What broke, and why

• •Read line by line, both sidebar layouts broke about half the bullets — 81 of 174 and 80 of 174 survived. The reader goes straight across the page, so each line of the sidebar is glued to the line of the main column beside it, and a bullet that wraps comes out in two pieces with sidebar text in between. The same thing misread 13 employer names.
• •Read in stored order with the sidebar on the left, every bullet survived but the name did not. The file holds the whole sidebar first, so the name was no longer the first thing read, and it was misread on 36 of the 36 resumes.
• •Read in stored order with the sidebar on the right, the result was identical to one column — because these PDFs happen to store the main column first. A different design tool can store the two columns in either order, and nothing on the page shows which.

What the mix looks like

These are the first lines of the customer service example, with the sidebar on the left, exactly as the line-by-line reader returned them.

• •Marcus Bell
• •Contact
• •marcus.bell@example.com
• •(216) 555-0148 Summary
• •Cleveland, OH Support specialist with five years in high-volume retail and SaaS queues. Most
• •recently raised first-contact resolution from 61% to 78% while handling 70 contacts
• •Education
• •a day.

The phone number and the word “Summary” share a line, the city runs straight into the first sentence of the summary, and “Education” lands in the middle of a sentence about contacts a day. A person looking at the PDF sees none of it, which is why the problem survives so many proofreads.

What this measurement does not tell you

• •It tests two reading strategies, not any vendor's applicant tracking system. Those are private and configured per employer, and nobody outside them can run this test on them.
• •Every PDF came from one PDF library. A file from Word, Google Docs or a design tool can store its text in a different order, and the stored-order results would change with it.
• •Some parsers detect columns and read each one separately. Where one does, a sidebar costs nothing — but you do not get to choose which parser reads your file.
• •It did not test headers and footers, tables, text boxes or images, each of which some parsers handle differently again.

What to do with it

If you want the result that holds across every reader, use one column: it was the only layout that came through both readers intact on all 36 resumes. If you prefer a two-column design, check the exact file you are going to send rather than trusting the template's description of itself.

/check reads your PDF both ways, in your browser with nothing uploaded, and shows the lines where the two readings disagree. Every template in the builder is one column, for the reason above.

Keep reading