> ## Documentation Index
> Fetch the complete documentation index at: https://docs.quantoraresearch.com/llms.txt
> Use this file to discover all available pages before exploring further.

# Units and periods

> How values are denominated, and why the unit field is not optional reading.

The same indicator can be published in different units by different countries, and the
API reports what the publisher actually published rather than forcing a conversion that
would misstate the source.

Read `unit` on every row. It is the field most often skipped and the one that causes
the most damage when it is.

## Units

| Unit                                | Meaning                               | Example                |
| ----------------------------------- | ------------------------------------- | ---------------------- |
| `percent`                           | A rate, already in percentage points  | `32.1` means 32.1%     |
| `index`                             | An index level, base varies by series | `114.2`                |
| `usd`, `eur`, `try`                 | A currency amount                     | scale given by `scale` |
| `thousands`, `millions`, `billions` | A count at that scale                 | `227.0` thousands      |
| `persons`                           | A headcount, unscaled                 | `219000`               |
| `ratio`                             | A dimensionless ratio                 | `0.62`                 |
| `basis_points`                      | Hundredths of a percent               | `44` means 0.44%       |

<Warning>
  Never compare or average across units. An unemployment rate in `percent` and payrolls
  in `thousands` are both labour market indicators and are not commensurable. Pooling
  them produces a number with no meaning.
</Warning>

## Index levels are not comparable across countries

An index is defined against a base period chosen by its publisher. Two countries'
indices at 114 are not "the same"; they are each 14% above whatever their own base was.

Compare **changes** in index levels, never levels themselves.

## Periods

`periodStart` and `periodEnd` bound what the reading covers. `frequency` names the
cadence.

| Frequency   | Typical publication lag       |
| ----------- | ----------------------------- |
| `daily`     | Same or next day              |
| `weekly`    | Days                          |
| `monthly`   | Two to six weeks              |
| `quarterly` | Four to twelve weeks          |
| `annual`    | Months, sometimes over a year |

Periods are reported as the publisher defines them. A quarterly row ending
`2026-06-30` covers April to June, whatever the country's fiscal convention.

## Seasonal adjustment

Where a publisher offers both, the API prefers the seasonally adjusted series, because
that is what the same figure means when compared across months.

Mixing adjusted and unadjusted readings within one series produces artificial swings
that look like real economic movement. The API does not do this, and neither should a
pipeline that joins across sources.

## Revisions

Official statistics are revised. A figure fetched today may differ from the same
period fetched last month, and that is the publisher correcting itself, not the API
changing its mind.

For point in time work, store `releaseDate` alongside every value and reconstruct what
was known on a given date rather than assuming today's history was always today's
history.