If you built your resume in Canva, Novoresume, or Zety in the last two years, there is a decent chance recruiters have never actually read it. A vendor engine study published Oct. 6, 2026 found that 61% of first-time resume uploads fail parsing before any keyword is scored, and almost all of the damage comes from one design choice: side-by-side columns. The fix takes ten minutes. The explanation takes a little longer.
What the Oct. 6 study actually found
Resume Optimizer Pro's parser processed 3,382 first-time resume uploads between Aug. 30 and Oct. 6, 2026, and 61% failed to extract cleanly before keyword matching began. Of those, 59% came out with lines merged from side-by-side columns, and 1.7% returned zero readable text at all.
That is not a design critique. That is the applicant tracking system (ATS), the software that reads resumes into database fields, giving up on the file before a recruiter sees anything. The ATS does not throw an error. The recruiter does not get a popup. Your resume simply shows up with "Senior Product Manager" sitting in the skills field and "Python, SQL, Figma" sitting in the job-title field, and the search filter they ran at 9:14 a.m. does not return you.
Resume Optimizer Pro engine, 3,382 uploads between Aug. 30 and Oct. 6, 2026, before any keyword matching.
The 61% is a floor, not a ceiling. The study only counts a fail if lines visibly merged or no text came out. It does not count the resumes where text extracted "cleanly" but landed in the wrong field, which a separate benchmark captures as a drop from 96% field completeness on a single-column DOCX to 52% on a two-column Canva PDF. Real functional failure is almost certainly higher than 61%.
Why two columns break parsers
ATS parsers read a document top to bottom, left to right, one line at a time; a second column puts your job title and your skills list on the same extracted line, so the parser scrambles both. There is no AI in that step. There is a text extractor, usually Apache Tika or something similar, pulling raw character coordinates off the page and reassembling them into lines by their vertical position.
Here is what happens on a two-column PDF, step by step:
- The parser scans row by row across the whole page width.
- "Senior Product Manager" (left column, line 1) and "Python, SQL, Figma" (right column, line 1) share the same vertical coordinate.
- The parser emits one line:
Senior Product Manager Python, SQL, Figma. - The ATS field mapper sees a long string, cannot locate the job-title pattern, and dumps the whole thing into a generic text blob or splits it wrong.
- The recruiter's filter for "Senior Product Manager" misses you.
Canva compounds the problem because it embeds visual positioning coordinates rather than semantic HTML text streams, Workday strips the CSS visual formatting on upload, and Greenhouse merges adjacent text blocks because two-column PDFs lack explicit structural reading-order tags. Three different platforms, three different failure modes, same root cause.
The platform-by-platform damage
Workday fails two-column resumes 28.3% of the time and Oracle Taleo fails them 34.1% of the time, according to a 2026 ATSChecker benchmark. Those are the two ATS platforms most likely to be between you and a Fortune 500 recruiter in 2026.
| Metric | Figure | Source |
|---|---|---|
| First-time uploads failing parse (Aug 30 - Oct 6, 2026) | 61% | Resume Optimizer Pro, n=3,382 |
| Of which merged-line (column-caused) | 59% | same |
| Of which zero readable text | 1.7% | same |
| Two-column failure rate on Workday | 28.3% | ATSChecker 2026, n=267 |
| Two-column failure rate on Oracle Taleo | 34.1% | ATSChecker 2026, n=142 |
| Field completeness: single-column DOCX vs 2-col Canva PDF | 96% vs 52% | Resume Optimizer Pro, n=200 |
| Two-column score penalty | roughly -14 points | ATSChecker 2026 |
A separate Jobloo analysis of 1,000,000+ applications claims two-column Canva resumes fail Workday and Greenhouse extraction in 78.4% of submissions. That is self-reported vendor data, so treat it as an upper bound, but the direction of the arrow is the same across every independent study: columns lose content.
The counter-evidence, and why it does not save you
Enhancv's sample of 60 interview-stage resumes found 44 of them (73%) used a double-column layout, and that finding is real but channel-blind. Those hires went to Amazon, Metalab, PwC, FedEx, and Dell, which are companies that also source heavily through referrals, recruiter outreach, and LinkedIn InMail where no parser is ever involved.
The two-column resume succeeds in that sample because the ATS was bypassed, not because it survived one. A pretty PDF emailed directly to a hiring manager by a mutual contact never touches Apache Tika. A pretty PDF uploaded through myworkdayjobs.com does, and gets eaten.
The practical read:
- Portal upload (Workday, Greenhouse, Taleo, iCIMS, Lever, SuccessFactors): single column, every time.
- Human handoff (recruiter email, referral intro, warm InMail reply): designed PDF is fine, and arguably better.
- Hybrid (you upload, then a recruiter also pulls your file manually): single column still wins, because the recruiter searches the parsed database first.
Keep both versions. Most people who care about this already do.
The single-column fix you can ship today
Rebuild your resume as a single-column layout in Word or Google Docs, then export as both DOCX and PDF. In Word that is Layout tab, then Columns, then One; in Google Docs it is Format, then Columns, then the single-column icon. Redistribute content vertically instead of horizontally.
A checklist that works on the parsers above:
- One column, full page width. No sidebars. No "skills" rail on the left.
- Standard section headers: Experience, Education, Skills, Projects. The parser is looking for these exact words.
- Dates on the right of each role, formatted
Mon YYYY - Mon YYYY. Avoid date ranges inside bullet text. - Job title on its own line, then company and location on the next line. Do not combine them with pipes or bullets on the same line.
- No text boxes, no tables for layout, no header or footer for contact info (Workday skips those regions on a lot of templates).
- Fonts: Calibri, Arial, Helvetica, Georgia, or Times. Nothing exotic.
- Save as .docx first. Export .pdf from the .docx, not from a design tool.
- No photos, no icons, no colored shapes behind text.
Canva makes this hard on purpose. Its export menu offers JPG, PNG, PDF, SVG, MP4, GIF, and PPTX, with no Word export, so even if you pick an "ATS-friendly" template the gallery hands you, you cannot ship it in the format Workday actually prefers. If you have to use Canva for the designed version, rebuild the ATS version in Word from scratch. Do not try to convert the PDF.
This is also the exact work Refolk takes off you: paste your history and the posting, get a single-column resume back, correctly tagged, with the right section headers and date formatting for the parser on the other side. The design version stays in Canva for the human handoff. The parse version is the one that gets you into the search.
Who gets hit hardest
Designers, marketers, and creative-role candidates get hit hardest and are the least likely to change, because for them the resume is also a portfolio artifact. ATSChecker's 2026 study found marketing and creative roles score lowest at 51/100, largely because of heavier use of designed templates.
The two-version workflow is non-negotiable for these roles:
- A single-column DOCX for every portal upload (Workday, Greenhouse, Lever).
- A designed PDF for recruiter emails, warm intros, and the "attach your resume" field on your own portfolio site.
- A link to the portfolio inside the DOCX, as plain text, not as a button or icon.
If you are an engineer or analyst reading this and shrugging, do not. The ATSChecker data shows two-column failures on technical roles too, and the Workday 28.3% number is averaged across every function. Your "clean minimalist" sidebar template is the same structural problem as the designer's photo-and-icons template.
Why advice alone will not fix this
There are roughly 118,000 US recruiters and talent-acquisition professionals in Refolk's index, against roughly 5,500 US career coaches and resume writers: a 1-to-21 ratio where the gatekeeping side outnumbers the advice side by more than an order of magnitude. The fix cannot scale through coaching. It has to scale through tooling.
The gatekeepers outnumber the resume coaches 21 to 1. Advice cannot fix a parse failure at that ratio. Tooling can.
That is the structural reason the fix has to live inside the resume builder itself, not in a blog post a candidate reads once. A builder that writes the parsed file in a single-column layout by default, tailors it to each posting, and scores the match before submit removes the failure mode at the source. The design file is for the human. The parsed file is for the database. Both should ship from the same source of truth.
The 10-minute audit
If you are not rebuilding today, do this check on your current resume before your next application.
- Open your PDF in a plain text viewer (on Mac, run
pdftotext resume.pdf -in Terminal; on Windows, upload to a free text extractor). - Read the output top to bottom.
- If you see your skills, job titles, and dates jumbled on the same line, your resume is in the 61%.
- If sections like "Experience" or "Education" are missing from the extracted text entirely, your resume is in the 1.7% that extract zero readable content from parts of the page.
- If the output looks like what a human would read, you are fine. Keep the file.
Candidates who have been applying for weeks with no responses almost always find the extracted text looks nothing like the PDF they see. That is not the recruiter ghosting you. That is the parser never getting your name into the database in the first place.
The Oct. 6, 2026 study did not create a new problem. It put a number on a problem the industry has quietly been working around for years with referrals and InMail. If you have those channels, use them. If you are uploading cold, use one column.
FAQ
Is a two-column resume really dead, or is this overblown?
Dead for portal uploads, alive for human handoffs. The 61% parse-fail rate on first-time uploads and the 28.3% Workday failure rate on two-column layouts are both measured on resumes going through an ATS. The 73% success rate Enhancv reports is measured on resumes that reached a human through a non-portal channel. Keep both versions, and send the right one to the right channel.
Will converting my Canva PDF to Word fix the problem?
No. Canva embeds visual coordinates rather than semantic text, so when you convert its PDF to Word you get a document full of text boxes and absolute-positioned frames that parsers treat the same way they treat the original two-column PDF. Rebuild the resume in Word or Google Docs from scratch using a single-column layout. Keep the Canva file as your designed version for email.
What about "ATS-friendly" templates on Canva, Zety, or Novoresume?
Treat the label as marketing, not a parse test. Even Rezi warns that some templates still confuse ATS systems, and the top results on Canva's "ATS-friendly" search are often two-column layouts with sidebars and photos. If the template has a left rail, a photo, icons, or a skills bar graphic, it will lose content to Workday and Taleo regardless of what the gallery calls it.
Does the fix work for Workday, Greenhouse, and Taleo specifically?
Yes, and those are the three it matters most for. The ATSChecker 2026 study ranks Oracle Taleo (34.1% failure) and Workday (28.3%) as the two platforms with the highest two-column failure rates, and Greenhouse's text-block merging is the documented third. A single-column DOCX with standard section headers, standard fonts, and dates on the right of each role is the layout all three parsers handle most reliably.