I’m trying to figure out the best performing but most customer friendly replenishment flow process. 

The replenishment is for only 1 product in the product range, so it needs to be filtered based on a trigger of ordering that particular product. The next order can also be determined by the quantity of products the customer ordered. 

How can I A/B test 2 different approaches using both static delays and reminders against using predictive analytics and sending emails based on the next expected order date? 

The complexity makes it difficult to:

• Setup because of the various conditions and branches
• Measure the impact and compare performance

Should I do this in separate flows that are split by the random sample condition and update a property? Or try to do it all in one? 

I can’t quite envisage which is the best way to tackle this yet.