Pigment has functions and methodologies to help you add forecasts to your models. This article guides you through the options.

Background

Pigment's forecasting tools fall into three main approaches:

1. Growth rate hypotheses

2. Statistical forecasting functions

3. Machine Learning forecasting through the Predictions feature (coming soon)

Each method offers varying levels of sophistication, flexibility, and automation depending on the use case. Here is an overview of the three forecasting methods available in Pigment.

Growth hypotheses

The first approach to forecasting in Pigment uses a growth rate hypothesis to predict future months of a given Metric. For example, you can hypothesize that year-on-year growth will be 20%, and to model this you apply the hypothesized rate to month data from last year.

Using Growth Hypotheses to Forecast Revenue gives examples of the formulas and Metrics to create for this.

Statistical forecasting functions

The second approach is to leverage Pigment's modeling functions to create forecasts using statistical models, such as regression analysis and exponential smoothing.

Linear regression

For simple forecasting, the FORECAST_LINEAR function fits a straight line to your data using the ordinary least squares method for linear regressions. It returns a forecast of the following type:

This function requires just one mandatory argument and so is convenient for simple forecasting. It does not account for seasonality.

Seasonal linear regression

You can create a forecast comprising a simple linear regression with added seasonality, using the SEASONAL_LINEAR_REGRESSION function. Its methodology is similar to the FORECAST_LINEAR function, but has an additional required argument: length of seasonality. Defining this allows the model to produce forecasts of this kind:

Simple exponential smoothing

Exponential smoothing tools let modelers use data smoothing, trend smoothing, and seasonality smoothing factors to customize the forecast. If your data does not show specific trends or seasonality, the SIMPLE_EXPONENTIAL_SMOOTHING function allows you to model your forecast with one smoothing factor:

1. alpha—for data smoothing, used to control the model’s sensitivity to the most recent data point

Double exponential smoothing

If your data shows trends but does not show seasonality, the DOUBLE_EXPONENTIAL_SMOOTHING function allows you to model your forecast with two smoothing factors, the above and:

2. beta—for trend smoothing, used to control the model’s sensitivity to the most recent changes in the trend, allowing the model to adapt more or less quickly to shifts in the trend direction

Triple exponential smoothing

For data with seasonality, the triple exponential tool FORECAST_ETS allows modelers the above smoothing factors along with:

3. gamma—for seasonality smoothing, used to control the model’s sensitivity to the most recent seasonal values, allowing the model to adapt more or less quickly to changes in the seasonal pattern

Predictions (ML Forecasting) - COMING SOON

Pigment’s Prediction tool generates customized forecasts automatically for inclusion in your model. It leverages advanced statistical and Machine Learning algorithms to analyze your historical data and make predictions about future values. You can use it for a wide range of planning needs, including demand planning, financial forecasting, or revenue projections. Use it with any time-dependent Metric in your Pigment Workspace. See Predictions and ML Forecasting for more details.