[P] Benchmark: Using XGBoost vs. DistilBERT for detecting "Month 2 Tanking" in cold email infrastructure?

About this article

I have been experimenting with Heuristic-based Deliverability Intelligence to solve the "Month 2 Tanking" problem. The Data Science Challenge: Most tools use simple regex for "Spam words." My hypothesis is that Uniqueness Variance and Header Alignment (specifically the vector difference between "From" and "Return-Path") are much stronger predictors of shadow-banning. The Current Stack: Model: Currently using XGBoost with 14 custom features (Metadata + Content). Dataset: Labeled set of 5k emai...