> ## Documentation Index
> Fetch the complete documentation index at: https://docs.getprova.com/llms.txt
> Use this file to discover all available pages before exploring further.

# Bayesian SUR

> Learn how Provalytics uses Bayesian Seemingly Unrelated Regressions to model related outcomes jointly instead of treating each KPI as isolated.

## Overview

The econometric foundation of Provalytics is Bayesian Seemingly Unrelated Regressions.

That matters because marketing outcomes rarely move independently.

Brand search, sessions, engagement, store visits, and conversion behavior often respond together inside the same behavioral system.

## Why joint estimation matters

If related outcomes share correlated error structures, modeling them jointly can produce more efficient and more realistic estimates than treating each equation separately.

## Why Bayesian matters here

Bayesian estimation turns coefficients into distributions rather than pretending they are fixed and perfectly known.

This helps the framework:

* represent uncertainty directly
* stabilize estimation in sparse or noisy environments
* preserve interpretability while remaining operationally useful