I got rejected by Workday with a resume I was proud of. So I built a tool that shows you what the software actually reads.
I am a finance manager in Dubai, ACCA, starting an EMBA at London Business School this month. Not a developer.
The rejection: last year, and more than once before that. Senior finance roles, the kind I already do for a living, applied for with a two-column resume a designer friend had made for me. The rejections came back within minutes of applying, sometimes the same hour. No human reads a finance manager's resume in three minutes. When it happened for the fourth or fifth time I pasted the PDF into a plain text editor and saw what the software was seeing: my sidebar sat in the middle of my work history and my name was nowhere near the top.
What I built: a scanner that runs the same kind of plain-text extraction Workday-class systems run and shows you the result next to your original, names every failure (tables flattened, name not found, icon bullets turned into boxes, two-column scramble), then rebuilds the resume into a single column with your exact wording and shows page one free. The clean Word and PDF are $5, once. No subscription, no signup, file deleted in 60 seconds.
How: Next.js on Vercel, Stripe, Upstash for anonymous counters, the Anthropic API for extraction and optional bullet rewrites. I did not write the code by hand. I built it with Claude Code over evenings and weekends, about three weeks from idea to first sale in May 2026, and I still ship most days the same way. Running cost is about $50 a month, most of it Vercel Pro; each free preview costs me roughly one cent in model calls.
Numbers, all real, and the scan figures are published live on the site:
5,436 resumes scanned since 2 August. 42.3% had a table the parser flattened, 26.7% had a name it could not find.
About 70 paid orders from 44 people since May, average order just over $6. Small, but it is a finance manager's side project, not a funded startup.
Free scan to paid: about 1.5%. Since I started showing the rebuilt resume on screen before payment (this week), roughly half of visitors now look at their rebuilt resume; the conversion from there is what I am working on.
Distribution, honestly:
One Reddit post with a table of failure rates got 779K views and gave me the only traffic spike in four months. The same sub removed my next post.
Google organic and direct are 60% of sales. Bing sent one buyer this week.
Product Hunt and SaaSHub badges: nice backlinks, zero attributed sales.
ChatGPT started sending buyers this week, two in one day, both landing on a post comparing AI resume builders. I am now writing for assistants as much as for Google.
Hardest problem: I cannot log into Workday to see what it does. So I built the checks from failure patterns I could reproduce (copy-paste order, glued tokens, missing headings, dates in tables), tested them against public resume templates, and published the method so anyone can call me on it. I say "Workday-class extraction", not "we replicate Workday", because that is what is true.
What I want from you: tell me if $5 is too cheap to be trusted. Several people have asked "what is the catch" and I do not have a good answer beyond "there isn't one".
