Bellwether is a revenue-weighted feedback engine that joins support tickets, store reviews, sales calls, and NPS to CRM accounts, then stress-tests rankings on 6,693 synthetic signals. The analysis reveals why mention counts are biased, how revenue weighting re-ranks themes, and the limitations of synthetic data. In a previous analysis, the author argued SaaS PM is ideal for data-driven professionals.
This article builds on that by demonstrating how to join implicit feedback with explicit account value. The tool ingests and clusters known themes around a fictional field-service SaaS company, Volta, with about €19.4M ARR across 10 named accounts and 214 SMB plus 1,890 self-serve accounts. Feedback comprises 6,693 signals from four sources.
The ranking defaults to volume: offline mode 1,247 mentions, dark mode 903. Revenue weighting prioritizes accounts renewing soon. The four sources carry distinct biases: support tickets dominate SMB volume, reviews favor self-serve, sales calls represent enterprise, and NPS captures promoters.
A key insight is that the needed revenue data resides in the CRM, not the feedback tool. The code powering the analysis is public on GitHub. The article explores why mention counts bias roadmaps, how one click re-ranks the same ten themes, and whether the ranking survives its own assumptions, alongside three harder clustering problems.
Source: Towards Data Science · Summarized by HeadlinesBriefing