Partner ACV Forecast Engine

Pipeline forecasts treat deals as independent events. They are not — quarter-end pressure, macro conditions, and shared activation capacity make them move together. Assuming independence produces a forecast that looks confident and understates real risk in both directions.

A three-file Python package pairing two independent estimators. A factor-structured Gaussian copula Monte Carlo injects dependence through a shared market factor plus segment and activation-capacity factors, preserving each deal’s marginal close probability exactly and guaranteeing a positive semi-definite correlation matrix by construction. Alongside it, a conjugate Normal-Inverse-Gamma Bayesian run-rate model with trend and Fourier seasonality produces Student-t posterior predictive intervals that widen honestly as the horizon extends. Three deal states with correlated close-by-activate gates, a tornado chart ranking which single deal most moves the probability of hitting target, and a documented Close CRM REST adapter.

Replaced a single-number forecast with two independently derived distributions and an explicit read on where they disagree. The tornado output changed which deal got worked: unsticking one stalled activation moved the probability of hitting target more than closing either of the two largest opportunities, which were already priced in.