> ## Documentation Index
> Fetch the complete documentation index at: https://confidence-auth-testing.mintlify.site/llms.txt
> Use this file to discover all available pages before exploring further.

# Metric Data Errors

> When no data is available for metrics or metric results show errors

## Symptoms

You're experiencing this if:

* Exposure calculated correctly (non-zero exposure available for the
  experiment)
* Metric calculation job completed successfully
* Metric results delivered check shows an error

## Common Causes and Diagnoses

### Cause 1: Fact Table Returns No Data

#### Explanation

If your fact table query returns 0 rows, the problem is in your fact table
query.

#### Diagnosis

**1. Find your fact table query**

1. Navigate to the Design tab of your experiment
2. Go to the metric you want to troubleshoot
3. Select the **Fact table**
4. Copy the fact table query

**2. Run your fact table query**

1. Paste the fact table query into your data warehouse query editor (Warehouse Native only)
2. Run the query
3. If the query returns 0 rows, the problem is in your fact table query

If it returns data, add a filter to your query to only include facts for the
period your metric is calculating for to verify that there is data also for the
relevant period.

### Cause 2: Mismatch Entity IDs Between Fact Table and Assignment Table

#### Explanation

If your fact table query returns data but the entity IDs in the fact table and
the assignment table don't match, Confidence cannot find the exposed users in the
fact table.

#### Diagnosis

If your fact table query returns data but the metric shows `No data`, a possible
reason is that the entity IDs in the fact table and the assignment table don't
match. For example, if the fact table and assignment table use the same user IDs
but the user IDs are in different formats, or truncated differently. Investigate
by comparing the entity IDs in the fact table with the entity IDs in the
assignment table.

### Cause 3: Empty Partitions Committed

#### Explanation

Confidence waits for data to arrive in fact table partitions. If the data
delivery cadence is set to **Data is partitioned hourly** or **Data is
partitioned daily**, partitions may be committed empty depending on your
settings.

#### Diagnosis

**1. Check data delivery cadence**

1. Navigate to the Design tab of your experiment
2. Select the fact table in the **Metric config** section
3. Check **Data delivery cadence** setting to see if it is **Daily** or
   **Hourly**

**2. Check max zero rows duration**

1. On the fact table configuration page, select the edit icon in the
   **Configuration** section on the right sidebar
2. If the **Disable empty partitions** option is enabled, Confidence never
   commits empty partitions.
3. If the **Disable empty partitions** option is disabled, Confidence commits
   empty partitions if data doesn't arrive within the configured window defined
   by the **Max zero rows duration** setting.

#### Example

Example scenario for a daily partitioned fact table:

```text theme={null}
- Sunday: Internal data pipeline for your fact table doesn't run
- Monday 9 AM: Partition commits empty (max wait time reached)
- Monday 11 AM: Weekend data finally arrives
- Result: Weekend partition already committed empty, data ignored
```

### Cause 4: Metric Has No Variance

#### Explanation

If metrics have data, but all the data is the same value, the metric has no
variance. This leads to the Metric data delivered check showing an error,
because Confidence is unable to use the metric for statistical analysis.

#### Diagnosis

**1. Check metric results delivered check**

1. Navigate to the Results tab of your experiment
2. Check the Metric results delivered check to see if it shows an error and if
   the error is that the metric has no variance.
3. Click on the metric's calculation status in the **Metric calculations**
   section on the right sidebar to see the details of the calculation.
4. In the logs section, expand a calculation entry to view the summary
   statistics for the metric. Here you can see the variance of the metric as well
   as the mean and the count.

A metric can have no variance because while there is some data for the metric,
all the data is the same value. For example, every user had `purchase_amount`
equal to 0 for the calculation period. Equal values are more likely when there are
only a few users in the calculation.

The issue can also occur if the metric is misconfigured. For example, if you
configured the metric to use thresholds and no users met the threshold, or if the
metric uses filters that are not set up correctly.

## Related Resources

> **[Metrics Introduction](/docs/capabilities/metrics/overview)**
>
> Overview of metrics in Confidence

> **[Fact Tables](/docs/capabilities/metrics/fact-tables.md)**
>
> Configure fact tables

> **[Exposure Calculation Stuck](/docs/troubleshooting/experiments/exposure-calculation-stuck.md)**
>
> Troubleshoot exposure issues