- Take data from marketing events and other sources and score prospects by chance of won
- understend what features are the most important to make portret of Ideal Customer Profile (ICP)
marketing channel, campaign, refferer or whatever
page category, some part of page path and exactly the page (for populer page)
scoring should handle with 20k — 100k uniq features
table structure: — prospect_id — features list (comma separated)
Important: Features shouldn't be an outcome of the goal completion event (for example: agreement amount)
table structure: — prospect_id — is proving started (boolean) — is funnel end event (boolean)
Question: Why so many columns if it could be just "prospect_id" with goal completion? Answer: BC it's straightforward to make a mistake here. With explicit fields it should be less often problems like these:
- wrong sampling (when we have additional subset of prospects that not a part of funnel, we do this via: "is proving started" (boolean)
- causation missing. when goal completion were before the "is proving started" event
example of dataset: https://docs.google.com/spreadsheets/d/1PY3ee3V6hUlAyhj2kZz8-xiU8V5iyvnLECbtse-zRsQ/edit?usp=drive_link