CVSharkHow the score actually works
Getting past Australian hiring software isn't about stuffing in keywords. Systems like PageUp, JobAdder and Workday read a resume in two passes, and the number CVShark gives you is built to reflect both. Here's the whole model, no black box.
Two passes, not one
Layer 1 — Deterministic parsing
Rule-based extraction that pulls your resume into database fields. It reads top-to-bottom and cares about structure.
- ✓ Linear, single-column text
- ✓ Standard section headings it recognises
- ✗ Trips on multi-column and tables
Layer 2 — Semantic ranking
The modern AI layer. It judges whether your experience actually fits the role, not just whether the words appear.
- ✓ Context and seniority, not just terms
- ✓ Catches keyword stuffing and hidden text
- ✓ Rewards quantified, outcome-led bullets
The scoring engine, live
Your composite is a weighted blend of four pillars, minus penalties for anything that looks like gaming the system. Drag the sliders to see how each one pulls the number, and watch the shape of a resume across all four at once.
score = clamp(0..100, 0.25·structure + 0.30·keywords + 0.30·semantic + 0.15·AU − penalties)
Solid start, a few tweaks and this could be strong.
This calculator runs the exact formula CVShark uses on a real resume. On the real thing, the four pillar scores come from an actual check of your resume against the job ad, not sliders.
Why Australia-specific
The AU pillar is 15% of the score, and it's where US-built resume tools quietly cost you marks. Three things they miss:
Paper geometry
Australian resumes target A4 (210 x 297 mm). US Letter can cause rendering glitches in the preview panes of systems like PageUp and JobAdder.
Contact formatting
Suburb, state and postcode only. A full street address reads as a privacy risk here, and wastes space a recruiter would rather see used on your experience.
Spelling
Australian spelling (organise, analysed, optimise) matters. Localised keyword dictionaries are looking for the Commonwealth spelling, not the US one.
| Dimension | US standard | Australian standard |
|---|---|---|
| Work rights | Rarely stated | Expected, stated plainly (citizen, PR, or visa) |
| Personal photo | Discouraged, varies | Left off (bias / Fair Work reasons) |
| Resume length | One page, strict | Two pages is normal and expected |
What a high-semantic bullet looks like
Semantic fit rewards Action Verb + Context + Quantified Outcome. Same job, two very different scores:
Responsible for managing client projects and teams.
Led 12 engineers across 6 enterprise cloud deployments in APAC, delivering on time and cutting system downtime 30%.
Whose score this is
This is our model, not the employer’s. There is no matching percentage sitting inside an applicant tracking system waiting to be beaten, and any tool implying otherwise is overselling what it can see.
What the score is genuinely good for is checking whether your resume speaks the ad’s language before a recruiter searches for it, and whether the document will survive being parsed into a database. Both matter. Neither is a gate a machine opens. What actually happens to your resume after you apply sets out what these systems really do, with the vendors’ own documentation.
See your real score
The simulator is the model. Run it on your actual resume and job ad, free.