Prophet — Decomposable Time Series Forecasting
Also known as: Prophet, Facebook Prophet, Meta Prophet, forecasting at scale, Prophet — Facebook/Meta Zaman Serisi Modeli
Prophet is a Bayesian structural time series model introduced by Taylor and Letham at Facebook/Meta in 2018. It forecasts a continuous series by decomposing it into separate, interpretable trend, seasonality, and holiday components, and is designed to be approachable for analysts working at scale.
Key highlights
- Decomposes the forecast into separate, interpretable trend, seasonality, and holiday components.
- Robust to missing data and outliers, and tolerant of trend changepoints.
- Practitioner-friendly: produces reasonable forecasts at scale with little manual tuning.
Intuition
This section is available to Pro members. Upgrade to Pro
How it works
This section is available to Pro members. Upgrade to Pro
When to use it
Prophet fits a single continuous time series with at least about 30 observations, especially business and operational data that show strong daily, weekly, or yearly seasonality plus calendar effects such as holidays. It assumes the series can be decomposed into trend, seasonality, and holiday components, tolerates trend breaks, and is robust to missing data and outliers. It is a good first choice when you want fast, interpretable forecasts without heavy parameter tuning.
Strengths & limitations
- Decomposes the forecast into separate, interpretable trend, seasonality, and holiday components.
- Robust to missing data and outliers, and tolerant of trend changepoints.
- Practitioner-friendly: produces reasonable forecasts at scale with little manual tuning.
- Designed for a single univariate series; it does not natively model relationships between multiple series.
- Best suited to data with clear seasonal and calendar structure; it offers little advantage on short or purely irregular series.
- As a flexible curve-fitting model it can over- or under-shoot trend turns if changepoints are mis-specified.
Common pitfalls
This section is available to Pro members. Upgrade to Pro
Applications
This section is available to Pro members. Upgrade to Pro
Frequently asked
What does Prophet actually model?
It models a continuous series as the sum of a trend, a seasonal component, and a holiday/event component plus error. Each is a smooth, interpretable curve, so you can see how much of the forecast comes from each source.
When should I use multiplicative instead of additive seasonality?
Use multiplicative seasonality when the size of the seasonal swings grows as the overall level grows; use additive seasonality when the swings stay roughly constant in magnitude regardless of level.
How does Prophet handle missing data and outliers?
Because it fits smooth components by Bayesian curve fitting rather than relying on a strict recursive lag structure, Prophet tolerates gaps and is robust to occasional outliers, making it convenient for messy operational data.
Can Prophet forecast several related series at once?
Not natively — Prophet is built for a single univariate series. For relationships between multiple series consider multivariate or state space approaches such as a Bayesian VAR or a structural time series model.
Sources
- 1.Taylor, S. J. & Letham, B. (2018). Forecasting at Scale. The American Statistician, 72(1), 37-45.
- 2.Taylor, S. J. & Letham, B. (2017). Prophet: Forecasting at Scale [Software]. Facebook/Meta.
You have read it. What now?
Cite this page
ScholarGate. (2026, June 1). Prophet. ScholarGate. https://scholargate.app/econometrics/prophet-model