Numbers don’t get read. Sentences do.
Owners open a dashboard, see twelve widgets and close it. A summary written in sentences actually gets read.
- Healthcare context · endocrinology, Miami FL
- 42 rules · 6 domains
- Deterministic · no LLM
Play with combinations
Two critical signals to address this week. Revenue this month is $47k (+11% vs 3-mo avg of $43k), operating margin at 40%, weighted AR at 53 days against your 45-day target.
BlueShield FL AR at 71 days
$18,300 outstanding, 26 days beyond your 45-day threshold. Trigger collections protocol.
Single payer AR breach (escalation)
BlueShield FL: $18,300 @ 71 days. Escalate to payer relationship manager, not just billing.
Weighted days in AR above target
53 days vs 45-day target. Cash conversion is slow — each extra day of AR is roughly $1,573 stuck in the system.
No-shows spike on Thudays
24% no-show rate vs 14% baseline. Likely a scheduling-friction pattern (end-of-week, after-work) worth investigating.
BlueShield FL concentration warning
38% of revenue from a single payer (up from ~29% in Q1). Losing them would compress operating margin by roughly 23%.
BlueShield FL dominates accounts receivable
50% of all AR is tied up with BlueShield FL. Their payment cycle effectively sets your cash cycle.
Last 12 weeks
Where the friction is
The workflow, and how much of it is built here.
Solid steps run in this sandbox. Dashed steps are what a real implementation adds. Select any step to read what it does.
- 01Data
- 02Scenarios
- 03Rules engine
- 04Summary
Monthly snapshot
Built in this demoA hardcoded synthetic dataset: one fictional endocrinology practice, one month of operation.
In detail
- →Generates an executive summary from 42 rules across six operational domains (revenue, AR, ops, retention, concentration, margin).
- →Recalculates the summary in real time as you toggle any combination of 5 scenarios — order-independent composition.
- →Surfaces second-order risks (payer concentration, AR aging) that single-metric dashboards usually miss.
- →Attaches estimated dollar impact next to each actionable bullet.
- →Hardcoded synthetic dataset (one fictional Miami endocrinology practice, one month).
- →TypeScript rules engine — each rule has a trigger predicate, severity, narrative template, and optional dollar formula.
- →Scenario toggles are pure functions that mutate the snapshot; the engine re-runs against the mutated state.
- →Charts via Recharts, client-only rendering to avoid SSR hydration mismatches.
- →No LLM calls. No backend. Fully client-side, deterministic, replayable.
- →Live ingestion from a PMS/EHR (DrChrono, Athena, NextGen) and 837/835 claim file parsing for real AR aging.
- →3–6 months of historical baselines per practice. Without baselines, deltas are meaningless.
- →Per-practice rule calibration — a solo cardiologist's thresholds aren't a 5-doc OB-GYN group's.
- →Push delivery layer (email, SMS, Slack) so the summary arrives in the inbox rather than waiting to be pulled.
- →Audit log + rule versioning so a practice manager can trace why a number flagged six weeks later.
- →HIPAA-grade infra for US deployment — BAAs with every vendor, encryption at rest, access logs, breach-notification flows.
The pattern transfers cleanly to any ops domain — dental labs, retail, logistics, on-demand services. The work is in the per-domain rule set, not the engine.