Mining Salesforce opportunity history with pandas to expose real sales processes before automating

The article argues that companies should examine how their sales process actually runs before handing it to an AI agent, since documented workflows often differ from practice. It proposes using Salesforce's OpportunityHistory table, which logs every stage, amount, or close-date change, as a ready-made event log. Using pandas, the author shows how to extract stage sequences, count distinct process variants, flag backward moves and skipped stages, and measure waiting times between stages. The analysis also reveals whether a few users are entering most updates on behalf of a team, which can signal off-system workflows. The author stresses that these metrics identify where to investigate rather than explaining why the gaps exist.