---
title: "Configure holidays | Grafana Cloud documentation"
description: "A holiday configuration defines dates or periods that can enhance the forecasted signal."
---

> For a curated documentation index, see [llms.txt](/llms.txt). For the complete documentation index, see [llms-full.txt](/llms-full.txt).

# Configure holidays

Forecast models automatically detect common seasonal patterns such as daily and weekly cycles. However, some seasonal or external events require additional configuration to be modeled accurately.

If these events are not explicitly configured, the model can incorporate temporary spikes into the baseline or fail to properly account for them during prediction periods. This can result in biased forecasts, misleading predictions, and inaccurate alerts.

[A screenshot of a misleading growing forecast caused by unhandled seasonal events.](/media/docs/grafana-cloud/machine-learning/screenshot-example-of-innacurate-forecast.png)

To improve forecast accuracy, configure holidays so the model can recognize these patterns.

A holiday configuration defines dates or periods that can influence the forecasted signal. Holidays are not limited to public holidays, and they should represent any scheduled or expected event that affects the data.

Expand table

| Event type                   | Example                          | Occurrences                                   |
|------------------------------|----------------------------------|-----------------------------------------------|
| Scheduled operational events | Data backups or product launches | Last day of each month at 02:00 UTC           |
| Marketing events             | Ad campaigns                     | Fourth Friday of November, first week of July |
| Calendar-based events        | Easter or bank holidays          | Good Friday, Easter Monday, December 25       |

## Create a Holiday

1. In the Grafana Cloud main menu, select **AI & machine mearning**.
2. From the AI & machine learning homepage, navigate to the **Holidays** tab and click **+ Holiday**.
3. Give your holiday a name and a description.
4. Choose whether to manually add holiday occurrences or use an iCal calendar URL.
   
   - Choose **Manual**
     
     Provide a **Holiday occurrence name** along with start and end dates, and then click **+ Add Holiday**. Repeat this step for each occurrence in your group.
   - Choose **iCal URL** and provide a publicly accessible iCal URL.
     
     This option allows holidays to be managed and updated from your preferred calendar tool. All events from the iCal calendar are included in the holiday.
     
     Events can have specific start and end times, or be configured as **All day**, which lasts 24 hours starting at midnight in the selected time zone.
     
     > Note
     > 
     > Any updates to the iCal are applied when the forecast model retrains each day.
5. Select the time zone to use for all day events.
6. Optional: [Link to an existing forecast](#link-holidays-to-a-forecast) for which this holiday is relevant to ensure that the predictions from those forecasts account for the new holiday.
7. Click **Save**.

## Link Holidays to a Forecast

To use a holiday in a forecast, link them together when creating or editing the forecast.

1. In Step 2 of the forecast creation or edit page ([Preview and tune the forecast](/docs/grafana-cloud/machine-learning/machine-learning/forecasting/#preview-and-tune-the-forecast)), expand the **Holidays** section.
2. From the dropdown, select the holidays to link to the forecast.
3. Continue creating or updating your forecast as usual.

The forecast now accounts for occurrences of the linked holiday during training and prediction.

## FAQ

#### My forecast has a holiday linked, but the future predictions for occurrence times are unaffected

Ensure that your holiday has some occurrences in the past and within the forecast’s training window. For your forecast to account for the linked holiday, the model must have ‘seen’ some occurrences in the training data. If your holiday only contains future occurrences or occurrences earlier than your forecast’s training window, then future predictions will be unaffected.

Suppose this isn’t possible because there are no past occurrences. In that case, the forecast will automatically begin to account for future occurrences correctly once the first event has been ‘passed’ in time and is within the training data.