# Computing Careers 2026 — data dictionary

Edition 1.0.1. Generated by the build from the files in this directory, so every figure below describes the data you have actually downloaded.

Licensed CC BY 4.0. See LICENCE.txt. You do not have to buy the book to check the book.

## The one thing to read before you judge the coverage

`sources.csv` carries a locator on 57 of 200 rows. Another 83 record why no locator exists. The remaining 60 have neither.

**Those 60 rows are grade 4 and 5, and only grade 4 and 5.** Grade 4 is an aggregator — a salary site built from what people said about themselves. Grade 5 is a company describing the thing it sells. They are in the bibliography because the book argues with them, not because it rests on them, and several are not documents at all: the publisher column says so in words like "widely circulated industry claim".

Coverage of a locator is therefore a function of grade, by design. Here it is, per grade:

| grade | rows | with a locator | with a stated reason | date read |
| ----- | ---: | -------------: | -------------------: | --------: |
| 1 | 75 | 42 (56%) | 33 | 75 (100%) |
| 2 | 33 | 14 (42%) | 19 | 6 (18%) |
| 3 | 29 | 1 (3%) | 28 | 14 (48%) |
| 4 | 50 | 0 (0%) | 3 | 1 (2%) |
| 5 | 13 | 0 (0%) | 0 | 0 (0%) |

`retrieved` follows the same logic: you record the day you read something you can go and read. It is close to complete at grade 1 and close to empty at grade 5.

## sources.csv

200 rows, 14 columns. The same content is in sources.json.

| column | filled | meaning |
| ------ | -----: | ------- |
| `id` | 200 (100%) | Stable identifier, unique across the bibliography. |
| `title` | 200 (100%) | The title of the work, as the publisher gives it. |
| `publisher` | 200 (100%) | Who published it. For grades 4 and 5 this may name a class of source rather than one organisation, because what is being recorded is a claim in circulation. |
| `date` | 200 (100%) | When the source was published, ISO 8601, to the month. |
| `retrieved` | 96 (48%) | When it was read. See the note on coverage below. |
| `grade` | 200 (100%) | 1 to 5, best first. The whole method is Chapter 5, free at /part-one/the-source-hierarchy/. |
| `url` | 57 (28%) | Locator. See the note on coverage below. |
| `doi` | 1 (0%) | Locator, where the work has one. |
| `no_url` | 83 (42%) | Why there is no locator, where that has been established. |
| `key_figure` | 200 (100%) | The figure this source supports, in words. |
| `caveat` | 200 (100%) | What this source cannot tell you. Filled on every row. |
| `method` | 87 (44%) | How the source produced its figure, where stated. |
| `note` | 8 (4%) | Anything else worth recording. |
| `used_in` | 200 (100%) | The chapter that relies on it. |

## gaps.csv

154 rows, 6 columns. The same content is in gaps.json.

| column | filled | meaning |
| ------ | -----: | ------- |
| `chapter` | 154 (100%) | The chapter where the gap bites. |
| `profile` | 100 (65%) | The career it concerns, where it concerns one. |
| `kind` | 154 (100%) | dimension, chapter or snapshot — what sort of gap it is. |
| `what` | 154 (100%) | What is missing. |
| `note` | 154 (100%) | Why it is missing, and what was looked for. |
| `closeable` | 22 (14%) | `yes` where a published source exists and nobody has compiled it. EMPTY MEANS NO, not unknown. |

## role-index.csv

452 rows, 4 columns. The same content is in role-index.json.

| column | filled | meaning |
| ------ | -----: | ------- |
| `title` | 452 (100%) | A job title found advertised in computing. |
| `covered_by` | 452 (100%) | The career profile that covers this work, or the words `not profiled` where the book confirmed the title exists and did not profile it. |
| `chapter` | 452 (100%) | Where that profile sits in the book. |
| `status` | 452 (100%) | profiled, growing, stable, contracting or absorbed. |

### A warning about `closeable`

It reads `yes` on the rows where a published source exists and nobody has compiled it, and is **empty on the other 132**. Empty means no. It does not mean unknown. The book decides this for every gap; the blank is a formatting choice in the export rather than an absence of judgement, and it should be written out as `no` in a future edition.

## Citing this

> Source bibliography, gap register and role index, Computing Careers 2026, Grade One Press, edition 1.0.1.

Third-party material inside these files is credited where it comes from. O*NET, published by the United States Department of Labor, and ESCO, published by the European Commission, are used under CC BY 4.0. Neither body has reviewed or endorsed this book.

Corrections: corrections@gradeone.press. Every correction is free and reaches every buyer.