Skip to main content

Part of WatEase, India’s agentic AI commerce and marketing platform

MCMC (Markov Chain Monte Carlo)

A class of algorithms (Metropolis-Hastings, Gibbs, NUTS, HMC) that draw samples from a probability distribution by constructing a Markov chain whose stationary distribution is the target.

Last updated

Definition

MCMC sits inside Bayesian MMM as the engine that approximates the posterior over channel contributions. Typical implementations use 2,000 draws after 1,000 tuning steps per chain across 4 chains — the convention from PyMC and Stan tutorials, calibrated against marketing data's typical noise level. The output reads as a histogram of plausible per-channel coefficients; the mean is the point estimate, the 2.5% and 97.5% percentiles are the 95% credible interval. Convergence is diagnosed via R-hat (should be < 1.01) and effective sample size (should be > 400).

How it applies in India

Indian sales calendars have sharp festive peaks, such as Diwali and the wedding season, and an MCMC fit needs those weeks marked as seasonal, or it credits the festival to the ads that ran during it.

Frequently asked questions

Why MCMC and not just gradient descent?

MCMC samples FROM the posterior; gradient descent only finds the maximum a posteriori (MAP) point. For Bayesian MMM you want the full distribution — that's where the credible interval comes from.

Related terms

Try WatEase

Run your WhatsApp commerce on the platform built for India — Cloud API, GST invoices, UPI checkout, opt-in tracking, and 0% markup on WhatsApp messages (competitors add 20–25%). Pricing is on request — quoted by the WatEase team.

Get Started Today