> ## Documentation Index
> Fetch the complete documentation index at: https://docs.variable.global/llms.txt
> Use this file to discover all available pages before exploring further.
# Data quality rating
> How Variable scores data quality across five dimensions, following the PACT methodology
Variable scores the quality of every input and dataset using a **Data Quality Rating (DQR)** based on the [PACT Methodology v2.0](https://www.carbon-transparency.org/). The aggregate DQR gives you - and anyone consuming your data - a quick read on how confidently a footprint can be reported.
Lower is better. The best possible score is **1**, the worst is **3**.
## How the rating is shown
Each input row and each dataset surfaces a single aggregate DQR (e.g., `1.4`, `2.0`). The number is colored to make hotspots obvious:
| Aggregate DQR | Color | Read |
| :------------ | :------ | :------------------------------------- |
| ≤ 1.5 | Green | Good - suitable for external reporting |
| 1.5 - 2.5 | Default | Fair - usable, document assumptions |
| ≥ 2.5 | Red | Poor - improve before relying on it |
The aggregate is the average of the five dimension scores below. Dimensions left blank don't count toward the average - but they also reduce confidence, so fill them in where you can.
## The five dimensions
Each dimension is rated **Good (1)**, **Fair (2)**, or **Poor (3)**.
### Technological representativeness
How well the data represents the actual technology used.
| Score | Label | Meaning |
| :---- | :---- | :--------------------------------------------------- |
| 1 | Good | Same technology |
| 2 | Fair | Similar technology (based on secondary data sources) |
| 3 | Poor | Different or unknown technology |
### Temporal representativeness
How current the data is relative to your reporting period.
| Score | Label | Meaning |
| :---- | :---- | :-------------------- |
| 1 | Good | Same reporting year |
| 2 | Fair | Less than 5 years old |
| 3 | Poor | More than 5 years old |
### Geographical representativeness
How well the data's location matches yours.
| Score | Label | Meaning |
| :---- | :---- | :---------------------------------- |
| 1 | Good | Same country or country subdivision |
| 2 | Fair | Same region or subregion |
| 3 | Poor | Global or unknown |
### Completeness
How much of the underlying activity is actually covered by the data.
| Score | Label | Meaning |
| :---- | :---- | :-------------------------------------------------------------------------------------------------------------------------- |
| 1 | Good | Activity data collected for all relevant sites for the specified period |
| 2 | Fair | Activity data collected for less than 50% of sites for the specified period, or more than 50% of sites for a shorter period |
| 3 | Poor | Activity data collected for less than 50% of sites for a shorter time period, or unknown |
### Reliability
How the data was obtained.
| Score | Label | Meaning |
| :---- | :---- | :---------------------------------------- |
| 1 | Good | Measured activity data |
| 2 | Fair | Activity data partly based on assumptions |
| 3 | Poor | Financial data or non-qualified estimate |
## Primary vs secondary data
Alongside the DQR, Variable tracks the share of **primary** (specific, measured) versus **secondary** (averaged, modeled) data behind a footprint. You'll see this as two percentages that add to 100%:
* **Primary data %** - directly measured at your site or supplier
* **Secondary data %** - database, sector, or estimated values
The split is informational - it doesn't change the DQR - but it gives reviewers a sense of how much of the result rests on direct measurement.
## Document your ratings
Alongside the scores, you can record a free-text **data quality description** - why those ratings were chosen, which production years and background datasets the figures rest on, and any caveats a reviewer should know. EN 15804 expects this narrative "data quality discussion" next to the ratings.
Edit it in the data quality panel beside the dimension selectors, on both products and materials. Like the ratings themselves, it's read-only for externally synced materials. Over the public API and MCP it reads and writes as `dataQualityIndicators.description`.
## Where DQR shows up
| Surface | What you see |
| :------------------- | :--------------------------------------------------------------------------------------------------------------- |
| Input row in a model | Aggregate DQR for that input; hover to see all five dimension ratings plus the primary/secondary split |
| Dataset detail | All five dimension scores, plus primary/secondary split |
| Material element | The five dimension scores, editable inline beside the dataset passport |
| Product footprint | Aggregate DQR for the product, weighted across its inputs |
| PACT API export | The full `dqi` block (`technologicalDQR`, `temporalDQR`, `geographicalDQR`, `completenessDQR`, `reliabilityDQR`) |
## Improving your rating
Focus where it matters most:
1. **Find your hotspots.** Sort inputs by impact contribution.
2. **Check their DQR.** A high-impact input with DQR 2.5+ is your highest-leverage fix.
3. **Replace secondary with primary.** Supplier-specific datasets (or a verified EPD) typically move all five dimensions toward 1.
4. **Re-run.** The aggregate updates as soon as you re-assign the dataset.
Improving DQR on low-contribution inputs is rarely worth the effort. Always prioritize by share of total impact.
## Related topics
* [Environmental impact indicators](/docs/help/impact-indicators) - What impacts are measured
* [Assigning datasets](/docs/getting-started/assign-datasets) - How dataset choice drives DQR
* [PACT API](/pact-api/welcome) - Exposing DQR over the PACT exchange