What applicant tracking systems actually do, what they cannot read, and the five dimension self check. Explained in full below, and free to download, keep and teach from.
The single most repeated claim about applicant tracking systems, that they automatically discard around three quarters of applications before a human sees them, is not true, and the advice built on top of it has made a great many resumes worse. The fear produces keyword stuffing, invisible white text, and stripped down documents that no longer say anything.
What the software does is more mundane and more consequential. It parses a document into structured fields, name, titles, employers, dates, skills, and then ranks candidates by relevance against the language of the job post. Nobody is thrown away. People are sorted, and being sorted to the bottom of a long list has the same practical effect as rejection while being an entirely different problem to solve.
Which means the two things that actually cost interviews are vocabulary mismatch and thin evidence, not parser failure. Parser failure is real but rarer, and it is almost always caused by design choices made in pursuit of looking good.
Anything the parser cannot read effectively does not exist, no matter how good it looks.
Single column. Standard section headings. Simple bullets. Dates as MMM YYYY. Two column layouts and tables used for positioning are the common failure, because the parser reads across the page rather than down the column and returns an interleaved mess.
Contact details inside a header or footer. Text rendered as an image. Photos, charts and skill meters. Each of these is read by the software as nothing at all, and contact details in a header are the most expensive mistake on this list, because the document parses perfectly and then cannot be replied to.
Take the post and pull the terms it repeats. Aim for eight or more genuine matches, using the employer’s words rather than your own synonyms for them. This is not keyword stuffing. It is answering in the language the question was asked in, which is the same courtesy you would extend in the interview.
Count the bullets that contain a number or a named outcome, and divide by the total. Most resumes score badly here and improve fast, because the numbers usually exist and simply were not written down. A bullet with no number is a claim, and a claim competes poorly against a measurement.
Parse safety, keyword match, evidence density, structure, and the ten second readability test. Reverse chronological, with detail fading as the roles get older. If somebody cannot say what you are good at within ten seconds of looking, the ranking problem was never the software.
Responsible for managing social media channels and improving engagement across platforms.
Why it ranks lowNo number, no named platform, no outcome. Responsible for describes a job description rather than a person. The parser reads it perfectly and it still says almost nothing.
A measurementRan four brand channels and grew combined monthly reach from 40,000 to 190,000 in eleven months, at a cost per thousand of under two dollars.
What changedSame work. Named scope, a start figure, an end figure, a period and an efficiency number. It also now carries the vocabulary a marketing post would search on, without a single word being added for the machine’s benefit.
Sits alongside the Online Readiness Roadmap, where the question is how you appear to a system rather than a person. Worth doing before any job application in your final year, and worth redoing rather than editing, because editing a badly structured document preserves the structure.
If a resume has been sent out repeatedly without response, run the five dimension check before rewriting anything. The failure is usually locatable, and it is usually either a two column layout or contact details in a header. Both take five minutes to fix. Pairs with the STAR framework, which handles the conversation the resume is trying to earn.
The worksheet above is yours with no conditions attached. If an annotated edition would help, the one with a completed example in every field and the notes on where people get stuck, that is what this is for.
One message per instrument, nothing else. Students in the class do not need this, the annotated version is handed out in the session.
Everything here is original teaching material developed for university courses and corporate training in Seoul. Use it, adapt it, teach from it. When quoting or summarising, cite Matthew Clement, CareerComms, and link this page.
If it is useful in your classroom, it has done its job.