Projections let you extend any metric on a chart beyond today's date using Rivet's linear regression forecasting. Once enabled, projected values appear as dashed lines alongside your actual data, giving you a forward-looking view of performance without leaving your report.
Projections are available to customers who have purchased Rivet's Revenue Diagnostics - Forecast product.
Enabling projections on a chart
To turn on projections for metrics on a chart:
Go to the chart editor
Click into Chart settings
Turn on the toggle next to "Show projections"
When projections are enabled, the chart displays:
Solid lines for actual data up to the present date
Dashed lines for projected values, both from the past (based on the regression line) and into the future
Viewing future projections
To project into the future, choose a date range such as "Current year" that extends into the future. When you do this, you'll see actual data up until the present date, as well as projected data from the past and into the future.
How projections are calculated
Rivet uses a statistical method called linear regression to calculate projections. Using linear regression, Rivet plots your historical data and draws a best-fit line through it. The slope of that line is the forecast, and the line is extended forward to produce projected values.
Linear regression is deterministic, meaning the same data always produces the same projection. The trade-off is that it smooths out variance and is less precise for outlier patterns such as seasonality and staffing changes.
Projections are estimates that provide directional guidance rather than precise numbers. The further into the future the projection extends, the wider the potential margin of error.
How much historical data is used
The amount of historical data Rivet draws on to calculate a projection scales with the time range you're viewing.
In general, Rivet uses twice the length of your selected date range. So if you're viewing a 3 month window, projections are based on the prior 6 months of data. This scales up to a maximum of 2 years of historical data, and never less than 6 months regardless of the range selected.


