MN business health · evidence snapshot 11 Aug 2026

Where the subscription machine breaks.

One sourced comparator for payment health, retention, refunds, payback and cash quality. It shows what fails, what is unknown, and where the comparison itself is invalid.

Largest measured gap

Loading the evidence…

The score is a distance from transparent triage gates, never an industry percentile.

Provisional

Comparable now—

Failing gates—

Definition-locked—

Excluded as unknown—

Open source data

01 · Failure map

What fails most

Ranked only where an MN value and a documented gate exist. Asterisked values remain provisional because the external comparator or company perimeter is not fully matched.

Within gate Watch Fail Critical

How the gap index works

Zero means the value sits inside the healthy gate. Forty begins failure. Eighty begins critical. The index is deterministic distance from the visible thresholds below; it is not a percentile and does not estimate probability of company failure.

02 · Benchmark dossiers

Every judgment opens

Hover a term for the fast definition. Expand a dossier for the numerator, denominator, reference logic and direct sources.

03 · Payment leak

75% was the wrong grain

The denominator is now locked: July renewal invoices failed first attempt at 51.3%, or 50.7% after excluding intentional Radar blocks. Still critical—just no longer inflated by retries.

First attempt failed—— of — July renewal invoices

Later succeeded——

Unpaid at cutoff——

The contaminated view

— failed attempts

Retries multiply the numerator. The same affected invoice can appear as many as — times.

The decision grain

— affected invoices

— failed attempts per affected invoice on average. Count customers or invoices once before comparing.

Retry policy leadpreviously_declined_do_not_retry

Historical first failures—

Invoices retried anyway—

Additional failed attempts—

Eventually succeeded— · —%

In July, — invoices began with this reason; — were retried — more times; — recovered by cutoff. This supports a billing-owner policy test. It does not prove booked loss, intent, or a revenue-lift forecast.

Open the internal database assessment →

04 · Definition lock

No denominator, no benchmark

A metric can be numerically precise and still be unusable. These are the locks required before the ranking becomes decision-grade.

05 · Missing metrics

Measure. Do not pretend.

Public data cannot produce credible VPN-specific failure bands for these metrics. They stay visible and unranked until Stripe’s private peer layer or matched internal cohorts exist.

06 · Evidence register

The receipts

Primary filings and first-party platform reports. Company figures are anchors, not universal VPN percentiles.

07 · Data contract

Human page. Machine source.

One canonical JSON payload carries metrics, thresholds, provenance, units, freshness and privacy. The page renders it; it does not maintain a second truth.

Schema: kairos-business-health/v1
Freshness: Evidence snapshot 2026-08-11; current MN values marked by their own cutoff.
Privacy: Gated Kairos War Room. Contains internal aggregate business and billing figures; no customer-level rows.
Scoring: Transparent operating triage gates; never described as an industry percentile.
Canonical payload: benchmark-data.v20260811.json
Internal evidence: All-Hands Evidence Audit · Database Integrity Assessment · Benchmark research note