One of the distinctions I like to make about modern ‘Data Science’ is between ‘data science for decision support’ and ‘machine learning’. h
Basically speaking machine learning, which is often product-focused – is generally something like ‘there’s this problem in fraud, credit scoring’ and we need an automated and deployed system. You’ll often work super close with engineering teams on this.
Deploying Bayesian models is possible, but still a bit cumbersome. However, in my own experience, most of the value of applied Bayesian analysis (and related analysis) is in situations where it can be called ‘decision support’.
Decision Support: An executive or decision-maker needs to make a business decision, often of a strategic nature.
If you’re working for an online lender it might be ‘do we have enough capital to withstand shocks from the market’
If you’re a SaaS it might be what is our customer lifetime value, and what is it via cohort.
There are numerous other examples out there.
How can Bayesian Analysis help with this?
One of the most useful things that Bayesian analysis can do is to allow you to communicate uncertainty about strategic topics.
Bayesian analysis will help you explain to your executives how they can better made decisions, and allow you to convert your findings into dollars. For example see the Supply chain optimization example by Thomas and Ravin.
For example, you can create your own loss function that translates into some metric such as money, or you may have different customer lifetime value for certain customers and may want to include that in your model.
Want to learn more?
There are examples like this with notebooks and advice in my course. I have over 5 hours of screencast content and the opportunity to be part of a larger Bayesian learning community, which has nearly 500 members.
If you’re at a SaaS company or a fintech company, why aren’t you learning Bayesian analysis to provide a more in-depth analysis of strategic topics?