Module 2 of 13 in the AI Track, a free, self-directed curriculum for building AI that moves revenue.
The idea
You won’t build these from scratch often. You do have to recognize them and know the tradeoffs. Logistic regression is the mental model for lead scoring: the odds that something converts. Decision trees, random forests, and gradient boosting are the workhorses for tabular data, the rows-and-columns stuff. k-means groups similar records. k-NN finds the nearest ones.
Here’s the part people skip. For structured prediction like churn or conversion or scoring, a well-built boosted tree usually beats an LLM. It’s more accurate, it’s faster, and it costs a fraction of a cent to run. Knowing when not to reach for a language model is the actual skill.
Why it matters
Lead scoring, churn, prioritization. That’s RevOps. When someone asks “can’t AI just score our leads?”, the honest answer is yes, and the right tool is usually a cheap model that runs in milliseconds, not a chatbot. Being the person who says that is worth more than being the person who reaches for the expensive option.
Build this
A basic lead-scoring model in scikit-learn, on a sample CSV, that predicts who converts. Even a rough one makes you credible on the most common RevOps ask there is.
How to work through it
- StatQuest on logistic regression, decision trees, random forests, and boosting
- Skim chapters 1 through 7 of Géron for the concepts, not the code
- Train the lead-scoring model on a sample dataset
- Write up when a boosted tree beats an LLM, in your own words
Resources
- StatQuest playlists on regression, trees, and boosting
- Hands-On Machine Learning by Géron, chapters 1–7