Method
Sources and method
Every number in the app traces back to a licensed source, a collection time and a count of ads. This page is the long version of every ⓘ.
Illustrative data
Data sources
Where the numbers come from
- Job ad aggregate
Australian job ads (illustrative fixture)
Skill-level counts, entry-level share, work rights, work mode and advertised pay.
- Licence
- Sources and licences being finalised — see Data & safety
- Updated
- Daily, 05:00 AEST
- Last run
- 21 Sept 2026, 7:05 pm
Up to date
- Official statistics
JSA Internet Vacancy Index (fixture)
Occupation-level demand for roles, including the 12-month change.
- Licence
- CC BY 4.0 (to be re-checked)
- Updated
- Monthly
- Last run
- 21 Sept 2026, 7:05 pm
Running late — numbers may be a day old
- Practitioner experience
Coffee-chat debriefs
What people in the work said after coffee chats — only ever shown in aggregate.
- Licence
- Platform data, shown only in aggregate
- Updated
- Continuous
- Last run
- 21 Sept 2026, 7:05 pm
Up to date
Technology logos come from Simple Icons (CC0). They are trademarks of their owners and are used only to identify the technology.
Kinds of source
Three shapes, used everywhere
Next to every number is a small mark for the kind of source it comes from. The shape carries the meaning, so it still works without colour.
Official statistics
Square
Published by government agencies such as Jobs and Skills Australia. Counted at occupation level, revised from time to time, and slower to move.
Job ad aggregate
Circle
Counts from licensed job-ad data, collected daily. Fast and skill-level, but ads are not the same as jobs: some are duplicates, some are never filled.
Practitioner experience
Diamond
Experience from verified practitioners after coffee chats. It’s personal and qualitative — never a statistic about a market.
Filters
From filter to number
When we collect an ad we file it into exactly one cell, made of four things:
- 1PlaceThe most specific region we can tell, e.g. Parramatta rather than Sydney
- 2Work modeOn-site, hybrid or remote. Ads that don’t say count as on-site
- 3LevelGraduate, other entry-level, or everything else
- 4Work rightsRequires PR or citizenship, not stated, or sponsorship offered
Your filter is turned into the set of cells it covers. Places expand to everything inside them, and rules are combined so that an ad matching two rules is still counted once.
“On-site in Sydney or Melbourne, plus remote anywhere in Australia” covers every Sydney and Melbourne region for on-site ads, and every Australian region for remote ads — added up once each.
Shares use the same cells for the top and the bottom of the fraction, so a percentage always means “of ads matching your filter”. A few shares deliberately ignore one filter — entry-level share ignores the level filter, for example — and say so in their definition.
Sample threshold
At least 30 ads, or no number
If fewer than 30 ads sit behind a number, we show “not enough ads” instead of a value, draw no trend line and raise no signal. We suggest how to widen the filter instead.
Salary comparisons need more: the skill premium needs at least 60 ads that state a salary.
Extraction
Measures read from ad text
Some measures depend on how an ad is worded. We use written rules, test them on a hand-checked sample of ads, and publish the known error next to each one.
- Entry-level share
Text-derived
How it’s measuredTitle and text rules: graduate, grad program, junior, entry level or intern — and not senior, lead, principal or manager.
Known limitsAbout 1 in 10 ads is classified wrongly in our hand-checked sample.
- No PR stated
Text-derived
How it’s measuredText rules for phrases like "must have full working rights", "Australian citizen" or "PR or citizen".
Known limitsNot stated does not mean you can apply. Many employers expect working rights without saying so.
- Remote or hybrid
Text-derived
How it’s measuredWork-mode rules on text; ads that say nothing count as on-site.
Known limitsMany ads leave work mode out, so this is likely an undercount.
- Contract share
Text-derived
How it’s measuredContract type from the ad’s structured field, or from text when it is missing.
Known limitsBeing measured on our hand-checked sample.
- Skills per ad
Text-derived
How it’s measuredMedian count of distinct dictionary skills found in each ad.
Known limitsBeing measured on our hand-checked sample.
Rules are versioned. When a rule changes, we re-run it over past ads so trends don’t jump because of our own change.
Signals
When we flag a move
- More ads than 4 weeks ago
- Ads in the last four weeks are at least 25% and at least 10 ads higher than the four weeks before, with at least 30 ads behind it.
- Fewer ads than 4 weeks ago
- The same test in the other direction: at least 25% and 10 ads lower. Worded neutrally — a fall is information, not a warning.
- More entry-level openings
- The share of entry-level ads is at least 10 percentage points higher than eight weeks earlier, with at least 30 ads behind it.
Signals describe the market. They never rank people and never tell you what to learn.
Boundaries
What we never collect
- We never scrape LinkedIn, SEEK or any site whose terms forbid it. Job-ad data comes from licensed APIs and published statistics only.
- We never buy or build profiles of individual job seekers, and we never sell data. No one can pay to change what the market shows, and any conflict of interest is disclosed.
- We never show practitioner feedback about an individual — only aggregated, with enough responses to protect identities.
Personal data is stored in Australia, in line with the Privacy Act 1988 and the Australian Privacy Principles.