Reacting to data after the fact is costly — missed targets, late pivots, and strategies built on guesswork. Forecasting in Databox lets you get ahead of that by projecting your metrics' future performance using historical data, seasonality patterns, and machine-learning models. Whether you're setting realistic goals, planning budget allocation, or stress-testing a strategy, forecasts give you a data-backed view of where things are heading before you commit.
Forecasting is accessed directly from Metric Insights. Navigate to Metrics > Metric Insights, select the metric you want to forecast, and click the Forecast tab at the top left of the page.
The forecast chart displays several elements together:
- Solid line — actual historical values up to the current date
- Comparison series — values from the comparison period (e.g., same period 366 days ago), shown as a secondary line
- Forecast line — the projected values from the current date forward, shown as a dashed line
- Optimistic range — the upper confidence interval, shown as a shaded area above the forecast line
- Pessimistic range — the lower confidence interval, shown as a shaded area below the forecast line
The color of the confidence intervals reflects the metric's favorable trend direction. If an upward trend is favorable, the optimistic area is green and the pessimistic area is red; for metrics where a downward trend is favorable, the colors are reversed.
Below the chart, the data table shows Actual, Forecast, and Compare rows for each period, making it easy to review projected vs. historical values side by side.
Three summary cards above the chart give you a quick read on the forecast:
- Forecast for [period] — the projected end-of-period value based on the selected date range
- Previous period — the actual value from the equivalent prior period, with the percentage difference vs. the forecast
- Year to date — the cumulative actual value so far, with the percentage difference vs. the current forecast
Use the controls at the top of the page to define the forecast scope:
- Date range — sets the period being forecast. Click the selected range to open the date picker, where you can navigate between fiscal years, select a quarter (Q1–Q4), or set a custom end month using the End of drop-down. Click Apply to confirm.
- Graph by — controls the granularity of the chart (month, week, etc.).
- Compare to — overlays a historical series for context (e.g., Previous period, Same period last year).
Click the Methodology tab in the right-hand panel to select how forecast values are calculated:
- Databox Forecasting Methodology — combines time series and linear regression models. Forecast values update automatically when impacting metrics are added or their future values are changed.
- Linear regression — projects future values based on the linear trend of historical data.
- Simple statistical methods — uses basic statistical approaches for straightforward trend projection.
Enable seasonality if your metric exhibits recurring patterns within a year. Select the intervals that apply:
- Weekly — identifies patterns that repeat within a week; applicable only to forecasts with daily granularity.
- Monthly — identifies patterns that repeat within a month; available for weekly or daily granularity, but not monthly.
- Yearly — captures patterns that repeat within a year; applicable to monthly, weekly, or daily granularity, but not yearly.
Select Country to account for public holidays in a specific country when calculating the forecast. This helps reflect the real-world impact of holidays on your data.
Use the Breakdown by drop-down in the Visualize tab to split the forecast by a dimension — for example, by campaign or region. This lets you compare projected performance across segments.
Click + Add impacting metric below the data table to include metrics that are expected to influence the forecasted metric. Set the anticipated change for each impacting metric and observe how the forecast adjusts in response. This is useful for modeling scenarios — for example, estimating the impact of a planned budget increase on revenue.
The forecast confidence score indicates how reliable the calculated forecast is. It is displayed at the bottom of the page and takes into account:
- 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 selected forecasting parameters
A higher score reflects a forecast built on more consistent, longer-term historical data. When the score is low, the confidence score panel lists the specific factors contributing to the lower reliability — such as a short historical window, a long forecast period, or parameter choices that increase uncertainty — so you know exactly what to address.
Click + Save forecast in the top right corner to save the current forecast configuration. Saved forecasts can be accessed at any time from the Explore tab of Metric Insights by clicking Open all saved forecasts.
To share a forecast, click Copy URL to copy a direct link to the current forecast view.
FAQ
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, refer to the in-depth guide 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 is required to generate a forecast?
At least 12 months of historical data is required for the selected metric. Forecasting is not available for metrics with less than 12 months of data.
