@Jacob Moore stealing the analogy from a Custora post, imagine the Pareto distribution as being a "coin flip" if a customer will return or not, on a 0 to 1 scale. If it's above 0.5, we'll assume they're coming back. It is exponential because for most businesses, a large proportion of customers only purchase once. The mu parameter will control the severity of this curve, meaning that if your business has a high churn rate, the model will also predict most customers as churned.

So the dist is really saying "what is the likelihood that this customer will return ever again?"

Thanks for the question!

https://university.custora.com/for-marketers/clv/advanced/pareto-nbd