# Data Dictionary

## Main data files

The CSV and JSONL distributions contain the same 49,000 records and the same six fields.

| Field | Type | Definition |
|---|---|---|
| `domain` | string | Lowercase domain from the source snapshot. Every released value is unique within the sample. |
| `categories` | string | Complete ordered category assignment. Multiple labels are separated by the pipe character (`|`). |
| `primary_category` | string | First ordered label in `categories`. Used to define mutually exclusive sampling strata. |
| `popularity_band` | enum | Source-rank interval associated with the record. |
| `source_rank` | integer | Position of the row in the popularity-ordered source extract. It is not an external traffic estimate. |
| `classification_version` | string | Version recorded in the source classification row. |

## Popularity-band values

* `rank_1001_10000`
* `rank_10001_100000`
* `rank_100001_1000000`
* `rank_1000001_5000000`
* `rank_5000001_10000000`
* `rank_10000001_plus`

## Related summary files

`category-summary.csv` gives, for each category, `label_occurrences` (every appearance of the label) and `primary_category_domains` (domains whose first label it is). Because domains can have multiple categories, the sum of category occurrences is greater than the number of domains.

`dataset-analysis.json` contains the sampling allocations, per-stratum quotas and descriptive summaries for the release. Category occurrence counts use all labels; primary-category counts use only the first ordered label.

The taxonomy CSV and JSON define all 59 categories, including descriptions and example domains.