The performance of the submitted models will be evaluated using the Root [Mean Squared Error (MSE)](https://scikit-learn.org/1.5/modules/model_evaluation.html#mean-squared-error), computed across **all prediction targets and all hours** of the two-week test period.

In this competition, there are a total of **four target variables**:

* **Customer count** for **Company A**
* **Hourly sales** for **three product categories** of **Company B**

The final score is calculated as the **RMSE** over all observations in the test set:
![](https://www.googleapis.com/download/storage/v1/b/kaggle-user-content/o/inbox%2F22956516%2Ff34a31eb509bfdaf967b31abc78ca5ec%2FBildschirmfoto%202025-05-06%20um%2013.24.03.png?generation=1746530681835870&alt=media)

---

## Submission File
For each `id` in the test set, you must predict the corresponding value for either customer count or category-level sales. The submission file must include a header and follow the format shown below:

    id, target
    A_2024-09-02_15_customers	0.2 
    A_2024-09-02_16_customers	0.3
    A_2024-09-02_17_customers	0.1
    A_2024-09-02_18_customers	0.8 

Each `id` is the unique combination of company name (anonymized), date, hour of the day, and type of the target (customers or revenue). The target values are scaled to lie in the interval [0, 1] for data privacy.