> ## Documentation Index
> Fetch the complete documentation index at: https://docs.reelevant.com/llms.txt
> Use this file to discover all available pages before exploring further.
# Datagraph Explorer
> Visualise the Datagraph model and measure how well its relations actually join
## Overview
The **Explorer** page — *Explore your data model and the relations between datasources* — validates the Datagraph model and measures the real overlap of every relation. Open **DataHub → Datagraph → Explorer**.
## Before You Begin
A Datagraph Schema must exist. Without one, the Explorer shows *No schema found* and offers **Go to schema**. See [Datagraph Schema](/advanced-guide/datahub/datagraph/schema).
## Tabs
| Tab | Content |
| --------------- | ------------------------------------------------------------------------------------------------------------------------------------------------ |
| **Diagram** | An entity-relationship diagram of the model. Drag to pan, scroll with the modifier key to zoom, or use **Zoom in**, **Zoom out**, and **Reset**. |
| **Relations** | One row per relation column pair, with its **From** and **To** side and its **Join match** percentage. |
| **Datasources** | The Datasources used in the model, with their field count, relation count, and primary key marker (**PK**). |
## Validation Summary
The Explorer validates the model whenever it loads and reports three kinds of finding.
| Finding | Meaning |
| ----------------------------- | ------------------------------------------------------------------------------------------------------------- |
| **Structural issues** | The definition itself is invalid — for example an unknown referenced table or a duplicate column. |
| **Low join match** | The relation is structurally valid, but the values on both sides barely overlap. |
| **Invalid relation analysis** | The overlap could not be computed, usually because a Datasource could not be resolved or has no Live version. |
Fix structural issues first: relation analysis only runs once the definition is structurally valid.
## Join Match
For each relation column pair, the Datagraph counts the distinct values on both sides and compares them.
| Metric | Definition |
| ---------------------------- | ------------------------------------------------------------------------------- |
| **Matching distinct values** | The smaller of the two distinct counts. |
| **Match percentage** | Matching distinct values divided by the larger distinct count, as a percentage. |
| **Distinct values (from)** | Distinct values in the source column. |
| **Distinct values (to)** | Distinct values in the referenced column. |
Relations are then grouped, and the **Relations** tab can be filtered accordingly.
| Filter | Selection |
| ----------- | ------------------------------------------- |
| **All** | Every relation. |
| **Good** | Match percentage of 50% or above. |
| **Weak** | Match percentage below 50%. |
| **Invalid** | Relations whose analysis returned an error. |
A match percentage of `—` means the analysis could not run. Check that both Datasources have a Live version, then reload the Explorer.
## Inspecting a Join
Click **Inspect join match** on a relation to open the **Join match analysis** dialog — *How well the join keys overlap between the two datasources*. It reports the status (**Good**, **Weak**, **Invalid**, or **Unavailable**), the **Table**, **Column**, and **Datasource ID** on the **From** and **To** sides, and the join metrics above.
### Reading a Weak Result
A purchases table joined to a product catalogue on the product reference returns 12,000 distinct references on the purchases side and 400,000 on the catalogue side. The match percentage is 3%.
That is expected: only a small share of the catalogue has ever been bought. A weak result is a signal to interpret, not always a defect. Investigate when you expect the two sides to overlap closely, for example a customer identifier shared between a CRM Datasource and a website events Datasource.
| Symptom | Likely cause |
| -------------------------------------- | ---------------------------------------------------------------------------------------------- |
| Match percentage near 0% | The two columns hold different identifier formats, for example a raw ID versus a prefixed one. |
| Match percentage far below expectation | One side is stale, partially imported, or filtered at the source. |
| Analysis returns an error | The Datasource has no Live version, or it was deleted after the model was written. |
## What's Next?
Correct the relations flagged by the validation summary.
Query the validated model.
Check whether a Datasource import is stale.
Share the UI walkthrough with campaign teams.