Could the Long Tail Handle Itself?
The acquisition question answered whether FetchTrack was worth buying. It didn't answer how Pawsome would actually manage it.
Especially the hundreds of small, single-vet accounts that make up most of FetchTrack's customer base but least of its revenue. So I went further: built out tier definitions, five features' worth of telemetry, three separate health signals (not one blended score — that mistake's already been made once in this story), and a working triage dashboard a CSM could actually run hundreds of accounts through.
Try It
What To Look For
- Click "Clear" on a flagged account — the submit button won't unlock until you write a real sentence. No dropdown, no rubber stamp.
- Switch to "Team View" — a manager sees the pattern across the team before any individual CSM's record, and has to state a reason too. No exemption for rank.
- Expand "accounts behaving as expected" — most of the portfolio is invisible by default, on purpose. Attention goes where it's earned, not everywhere at once.
What This Doesn't Solve Yet
This prototype models the scores, the flags, and the accountability workflow. It doesn't yet define the actual numbers behind them — exact escalation windows, precise deviation thresholds, the specific day-counts that turn a flag into an escalation. Those were deliberately left as ranges and principles, not hard-coded values, because setting them responsibly requires real usage data this fictional company doesn't have. Guessing at false precision here would be the same mistake this whole project argues against.
A Note on AI Assistance
Claude (Anthropic) was used as a thinking partner throughout this project — walking through what each feature's telemetry would realistically look like, stress-testing the health-score design against its own stated principles, and auditing the final prototype for internal consistency (it caught, and we fixed, several places where the mockup contradicted its own logic before this went live).
All domain framing, tier definitions, and design decisions are my own. AI was used as a collaborative assistant, not a replacement for engineering judgment.