Credible Interval
A range that contains the true parameter value with a stated probability (e.g., 95%). The Bayesian counterpart to a frequentist confidence interval.
Last updated
Definition
A 95% credible interval means: given the model and the data, the parameter lies in this range with 95% probability. This is what most people THINK a frequentist confidence interval means — but a confidence interval actually has a more convoluted meaning ("if we repeated the experiment many times, 95% of the resulting intervals would contain the true value"). Credible intervals are operationally honest: a board member can read them at face value.
How it applies in India
Around festive peaks such as Diwali, spend and demand rise together, so a channel’s interval can come out wide; a wide interval there is the model being honest about what the data can separate.
An example
Two campaigns both show a return of three rupees per rupee. One has a 95% range of 2.8 to 3.2; the other, run for a fortnight on a small budget, has 0.9 to 5.1. They are not the same result: the first is something you can plan on, the second is still mostly a guess.
How WatEase uses it
WatEase shows uncertainty as a range wherever it can, although not as a Bayesian credible interval in the strict sense. The quarterly geo-lift test reports “Ads added about N% more orders, ±M%” from a 95% band on its odds ratio. The budget planner shows a ± band from 200 drawn spend curves. A response-curve chart shades a 95% band when the modelling engine supplies one. Why a range says more than last click’s single number is set out on WatEase’s research page.
Frequently asked questions
Is a WatEase ± band a credible interval?
Not in the strict Bayesian sense. The geo-lift band is a 95% confidence band on the measured lift, and the planner’s band comes from sampling plausible spend curves. Neither should be read as a guarantee.
Related terms
- Bayesian MMMMarketing Mix Modeling implemented with Bayesian inference (typically MCMC sampling), producing posterior distributions over channel contributions instead of point estimates.
- 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.
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