Work out prophet trend estimate instantly with clear inputs, formula shown and shareable results.
Prophet models trend as a piecewise linear function: growth is constant between changepoints and shifts by a fitted delta at each one. Projecting forward means taking the growth rate of the final segment and extrapolating it. That is why Prophet forecasts are so sensitive to a changepoint detected near the end of the series — the last segment's slope sets the entire future path.
Piecewise linear trend
g(t) = base + (growth x multiplier) x days since changepoint; counterfactual = base + growth x days
A changepoint fitted to the last few noisy observations can produce a steep final slope. Reduce changepoint_prior_scale or restrict changepoints to the first 80 percent of the history.
Whenever there is a natural ceiling such as market size or physical capacity. Linear growth will otherwise forecast past what is possible.