Reasons to go Bayesian Debunked

Be it Marketing Attribution or Experimentation, Bayesian methods are complex and hard to get right.

If you asked any Bayesian why they went Bayesian, they would tell you that "because it is more intuitive".

But Bayesian methods are hardly intuitive and most often require unnecessary complexity which leads to inaccurate results.

People are sold a very superficial version of Bayesian methodology - "you can update your beliefs as you get more evidence".

But this statement hides behind a lot of crucial details. To do Bayesian inference correctly one has to think deeply about the prior distributions and distribution convolutions.

However in reality, especially in MMM and AB testing, people just plaster half normal distribution or normal distribution as priors without any deep thought.

In reality, people end up sticking to a much more complex method just because they believe 'it is intuitive'.

There are other popular myths like:

📌 Bayesian statistics tell you what you want to know, frequentist ones do not.

📌 Frequentists don’t state their assumptions, Bayesians make the assumptions explicit

📌 Frequentist statistical tests require a fixed sample size (related to AB testing)

📌 Bayesian methods are immune to peeking at the data (related to AB testing)

📌 Bayesian inference leads to better communication of uncertainty than frequentist inference

All these myths are excellently addressed by Georgi Georgiev in his blog.

In addition you can also find link to articles on why we at Aryma Labs prefer Frequentist methods for MMM, Experimentations and Causal Inference.

---

Resources:

Adopting MMM for the first time? Use Frequentist MMM
https://open.substack.com/pub/arymalabs/p/adopting-mmm-for-the-first-time-use?r=2p7455&utm_campaign=post&utm_medium=web

Two key problems ailing Bayesian MMM
https://open.substack.com/pub/arymalabs/p/two-key-problems-that-ails-bayesian?r=2p7455&utm_campaign=post&utm_medium=web

Which method provides chances for greater manipulation - https://open.substack.com/pub/arymalabs/p/which-technique-provides-for-greater?r=2p7455&utm_campaign=post&utm_medium=web

Want performance guarantees ? choose Frequentist MMM - https://open.substack.com/pub/arymalabs/p/want-performance-guarantees-choose?r=2p7455&utm_campaign=post&utm_medium=web

Why you shouldn't use Geo tests to fix priors in Bayesian MMM.
https://open.substack.com/pub/arymalabs/p/why-you-shouldnt-use-geo-tests-to?r=2p7455&utm_campaign=post&utm_medium=web

Bayesian Marketing Mix Modeling's Stating the obvious problem
https://open.substack.com/pub/arymalabs/p/bayesian-marketing-mix-modelings?r=2p7455&utm_campaign=post&utm_medium=web

---

Thanks for reading.

For consulting and help with MMM implementation, Click here

Stay tuned for more articles on MMM.