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

# Thresholds

> Configure auto or constant thresholds to trigger alerts when a validation metric crosses a boundary.

Thresholds define when a validation result becomes an incident. Datachecks supports two types: **Auto** and **Constant**.

|                      | Auto                                   | Constant                                  |
| -------------------- | -------------------------------------- | ----------------------------------------- |
| **Setup**            | Automatic                              | Manual                                    |
| **Adapts over time** | Yes                                    | No                                        |
| **Best for**         | Dynamic data with natural fluctuations | Fixed compliance or operational standards |

## Auto thresholds

Auto thresholds use a machine-learning engine to establish baselines from your historical data. The system analyzes patterns — including seasonality and trends — and continuously updates its expected ranges as new data arrives.

**How it works:**

1. Datachecks runs preliminary statistical tests to understand the shape of your data series.
2. It learns the expected range based on historical values and the parameters you define.
3. Bounds are periodically refreshed to stay accurate as your data evolves.
4. When a new metric value falls outside the learned bounds, an incident is created.

Auto thresholds require no manual configuration and adapt automatically to changes in your data over time.

## Constant thresholds

Constant thresholds set a fixed boundary using a comparison operator and a value. An incident is triggered whenever the metric crosses that boundary.

**Syntax**

```
threshold: "<operator> <value>"
```

**Supported operators**

| Operator | Meaning                  |
| -------- | ------------------------ |
| `>`      | Greater than             |
| `>=`     | Greater than or equal to |
| `<`      | Less than                |
| `<=`     | Less than or equal to    |
| `=`      | Equal to                 |

**Examples**

```yaml theme={null}
# Alert when row count exceeds 30
threshold: "> 30"

# Alert when null percentage is 5% or more
threshold: ">= 5"

# Alert when freshness exceeds 24 hours (86400 seconds)
threshold: "> 86400"

# Alert when any duplicates exist
threshold: "> 0"

# Alert when row count between two sources differs
threshold: "< 1"
```

Constant thresholds are well-suited to metrics with predictable, well-understood boundaries. They do not adapt to fluctuations in your data over time.