Computing Careers
The first research project published here, on one corner of the subject. It asks what computing work actually pays, how people actually enter it, and which of those two questions anybody has bothered to measure.
The short answer to the last part is that for most of it nobody has, and the count is set out below. That finding is what shaped everything else.
What it asked
Ten questions, put to every career it examined, and put the same way each time: where the work exists, whether it can be done remotely, how people get in, what it pays, how many people are competing, what it costs to enter, what it costs to stay, how exposed it is to automation, what else could remove it, and what employers actually screen for.
Asking the same ten questions of unlike careers is what makes them comparable. Asking them against a fixed evidence bar is what makes the comparison worth anything, and it is also what produced the register of everything the questions could not be answered about.
What it produced
- Findings — single questions settled in public, with the working shown and every figure graded, listed with every other report under Reports
- Sources — 200 graded records, each with its date and what it cannot tell you
- Gaps — 154 questions that could not be answered at the required grade, 22 of them closeable by somebody with the right local knowledge
- Roles — 452 job titles, each mapped to the work it belongs to
- Registries — 11 publishers of graduate earnings matched to tax records, 17 consumer regulators with power over training providers, and the countries whose law makes an employer publish what it pays
- The dataset — the bibliography, the gap register and the role index as CSV and JSON, CC BY 4.0, free to copy and build on, and listed with everything else published free
- Computing Careers 2026 — the book it produced, and the one thing here that is sold
Everything except the book is free and needs no email address. The bibliography and the gap register can be reused commercially with credit, which means the evidence can be checked and argued with by anybody who never buys anything.
What it does not cover
Stated here rather than discovered later.
Most of the world. The evidence is good for the United States, fair across Western Europe, and thins out fast after that. Several careers carry a figure for one region and nothing anywhere else. Two countries have an annex of their own and 15 more need one.
How long entry takes, for almost anything. This is the first question every beginner asks and it has no measured answer for any of the computing work examined. All 46 careers declare it as a gap.
Whether the method picks well. Scoring ten dimensions is an argument that it beats looking at pay alone. Nobody has followed two groups of entrants for five years to find out, this research has not run that study, and until somebody does the framework is a well-organised argument rather than a tested instrument.
That last one is the largest open question about this work, and it is published alongside the work rather than left for a reader to notice.
Sending something back
The gaps marked closeable are the cheap ones: a published source exists and nobody here has compiled it. Several of them would be a twenty-minute job for somebody who reads the right language or knows the right national portal.
If you can close one, or if a figure here is wrong, the address is on the corrections page. A correction with a publisher, a date and a locator can be acted on. One without a source is still worth sending; it just takes longer.
Reports from this project
-
Finding
Which computing certifications can an individual not buy?
Duck Creek, Workday, Guidewire and Epic certify only the people who already work on their platforms. None publishes an exam price, because there is no purchase to make.
-
Finding
Professors are measured by rank. Software engineers by one occupation code.
A salary site gives three unrelated jobs the same figure to the dollar. Looking for evidence at the grain people actually work at, across 46 computing careers, found measured pay for 16 and a measured time to enter for none.
-
Finding
One school advertised 86% job placement. Its own metrics said 30%.
A placement rate is a fraction, and schools choose the bottom of it. A US regulator found BloomTech advertising 71% to 86% while internal metrics ran closer to 50%, on agreements capped at $30,000.
-
Finding
Is software development a fast-growing job in the United States?
The widely quoted 15% bundles two occupations. Taken apart, software developers are projected at 15.8% and QA analysts and testers at 10.0%, from the Bureau of Labor Statistics' own detailed table.
-
Finding
What does a tech certification cost to hold for ten years?
Between $165 and $2,225, computed from each issuer's own published exam fee and renewal terms. The advertised exam price is not the price, and there are only three renewal mechanisms in the whole market.