Help Center

Forecasts

Forecast your data

Learn how to forecast values for selected metrics and predict future developments for your business based on historical data.

Availability

Users, editors and admins

All accounts

Feature exclusive to select subscription plans

In Databox, you can predict your metrics' performance by leveraging historical data, seasonality patterns, and machine-learning models. This empowers you to make informed, data-driven decisions, helping you achieve your objectives and maintain a competitive edge.

Add a forecast

To add a new forecast, navigate to Forecasts > Saved Forecasts and click Add Forecast, or go directly to Forecasts > Forecast Explorer. In Forecast Explorer, you will need to:

Select the metric you want to forecast.
Define how far into the future to forecast values.
Choose the amount of historical data to include in the forecast model.

Note: Weekly granularity is only supported if Monday is set as the first day of the week in your account. Learn how to set your first day of the week here.

Once the forecast is generated, a forecast story explanation will be displayed above the chart.

The chart will be created with the following elements:

Historical data: Displayed as a solid blue line, this represents the historical data leading up to the current value.
Forecast line: Beginning at the current value and shown as a blue dashed line, this line indicates the most likely values for the forecast period.
Confidence interval: The interval provides a range of values within which the value is most likely to fall during the forecast period. It contains the optimistic and pessimistic forecast ranges.
- Optimistic forecast range: The area above the forecast line, it shows where the value is expected to land in best-case scenarios.
- Pessimistic forecast range: The red area below the forecast line, it shows where the value is expected to land in worst-case scenarios.

The color of the areas depend on the metric's favorable trend setting. If an upward trend is considered favorable, the optimistic area will be green and the pessimistic area will be red. Conversely, if a downward trend is considered favorable, the optimistic area will be red and the pessimistic area will be green.

For a visual guide to the individual elements and their meanings, click the How to Read a Forecast button located in the lower right corner of the chart.

Evaluate the accuracy of the forecast

The forecast confidence score on the Forecasts > Forecast Explorer page indicates the reliability of the calculated forecast.

This score considers several factors, including:

The amount of available historical data.
The length of the forecasted period.
The number of anomalies detected in the data.
The volatility of the available data.
The settings of the forecasting parameters.

A higher confidence score is achieved with a larger and more consistent amount of historical data for the metric.

Customize the forecast model parameters

In the forecast modeling parameters section, you can define seasonality and holiday options in more detail:

Seasonality: If the data for the selected metric exhibits recurring patterns on a weekly, monthly, or yearly basis, enable the include seasonality option. This will allow the model to detect and factor in these patterns during calculations.
Seasonality options depend on the granularity of your forecast:
- Yearly Seasonality: Captures patterns within a year. Applicable to forecasts with monthly, weekly, or daily granularity, but not yearly granularity.
- Monthly Seasonality: Identifies patterns within a month. Available for forecasts with weekly or daily granularity, but not monthly granularity.
- Weekly Seasonality: Identifies patterns within a week. Applicable only to forecasts with daily granularity.

Holidays: Enable the include holidays option if you want to account for public holidays in a specific country or financial market when calculating values. This helps ensure that forecasted values reflect the impact of these holidays.

Save a forecast

To save the forecast, click on the Add to Saved Forecasts link in the upper right corner of the page. This action will store your forecast on the Forecasts > Saved Forecasts page.

Clicking View details on any of the saved forecasts will take you to the Forecast Explorer page, where you can review and inspect the setup of the forecast.

Recalculate a forecast

In addition to saving a forecast, you can also recalculate an existing one. To do this, click the Recalculate Forecast link in the top right corner of the Forecast Explorer page while viewing a saved forecast.

Recalculation points will be marked as clickable green dots on the x-axis of chart. Click these dots to review the previously calculated values.

clickable-2

Frequently Asked Questions

Can a forecast be saved to the Saved Forecasts page multiple times?

Yes, a forecast can be saved to the Saved Forecasts page multiple times, but only if the setup of the forecast—such as the metric, historical data, forecast period, and modeling parameters—differs each time.

How are forecast values calculated?

Forecast values are calculated using a combination of historical data, seasonality patterns, and advanced machine-learning models. For a detailed explanation of the model and methodology used, please refer to the in-depth guide available here.

How can forecasts be valuable for my business?

Forecasts can provide significant value to your business by enabling you to make data-driven decisions, anticipate future trends, and plan strategically. To explore how forecasts can specifically benefit your business, learn more here.

How far into the future can values be forecasted?

Forecasted data can be displayed on the chart for up to 24 months.

How much historical data can be displayed on the forecast chart?

Historical data can be displayed on the chart for up to 24 months.

How much historical data is required to generate a forecast?

To generate a forecast, you must have at least 12 months of historical data for the chosen metric. Forecasting is not possible for metrics with less than 12 months of historical data.