of the displayed $11,628 July proxy result sits in paid tests: 78%. Paid intent, not recurring revenue.
Current hypothesis ranking · not a conclusion
Three survivors.
One lead.
What our current evidence supports, what it does not, and the cheapest way to find out.
01 · Two axes
Best evidence is not best asymmetry
The ranking keeps signal and upside apart. Conflating them creates false conviction.
Asymmetric upside →Signal strength →
Story riskDecision zoneDiscardOperational proof
Company-reported signal
Desk signal only
Click a candidate to jump to its test.
01 · Lead · internal
Paid-Test Conversion Strike
4/5 signal4/5 payoff skew2/5 copy resistance
Turn the existing July paid-test cohort into durable contracts—or learn, account by account, why it does not convert.
Mysterium owns the accounts, product context, telemetry and commercial conversations. The method is copyable. This cohort is not.
A small operating board and focused intervention can unlock recurring revenue—or produce a reusable map of why paid demand dies.
Strongest attackThis may be sales hygiene dressed as strategy.
The cohort may be one-off evaluations, low-fit buyers, or tests that cannot convert economically. Gross margin, delivery capacity, concentration, buyer access and loss reasons are missing. A prettier CRM view is not an opportunity.
Cheapest prove / kill
One row per paid test: buyer, use case, amount, success criterion, economic buyer, blocker, next action, owner, decision date, expected MRR, contribution margin and conversion probability. Resolve twice weekly.
Kill condition
No reachable economic buyer or repeat need; cohort cannot be reconciled; or expected retained contribution cannot justify the delivery and product work.
Cost / next decision
12–20 Kairos hours, excluding MN implementation. 2 working days after access: reconcile and rank. 2 weeks elapsed: conversion, contribution and reason-coded losses.
Expand evidence and limits
- Company-reported: July amounts and paid-test share in the 08-07 all-hands deck.
- Unknown: customer concentration, gross margin, product burden, repeat need and decision-maker access.
- Interpretation: the amount is sufficient to investigate, not sufficient to forecast.
02 · Challenger · internal
Payment-Recovery Torpedo
3/5 signal4/5 payoff skew1/5 copy resistance
Route eligible failed Stripe payments through the planned Adyen fallback and measure incremental retained revenue—not approval-rate theatre.
The deck reports Payments Orchestration live and instrumented; Adyen fallback is named as the next bet.
The edge is timing, implementation proximity and the actual failure stream—not a proprietary payments idea.
The test acts on demand that already exists. A small net recovery can repay the work; a null result can stop without a strategic hangover.
Strongest attackThis is payment plumbing, not a strategy.
Dual routing can lift approvals while worsening fees, fraud, chargebacks, latency, or low-quality retention. Without failure-class eligibility and a stable Stripe baseline, any uplift is uninterpretable.
Cheapest prove / kill
Pre-register eligible failure classes, baseline approval, routed cohort, incremental captured revenue, processing cost, fraud/chargeback guardrails, issuer and geo splits, and 30-day retained value.
Kill condition
Net recovered retained revenue is non-positive; uplift vanishes by failure class; or baseline and cohort cannot be made comparable.
Cost / next decision
4–8 Kairos hours for test design and analysis, excluding MN implementation. 1 working day with the payments owner; 1 billing cycle to scale by class or stop.
Expand evidence and limits
- Company-reported: live Payments Orchestration and planned Adyen fallback.
- Unknown: technically recoverable failure volume, net value, fraud profile and retained quality.
- Category risk: ordinary operational improvement can still have asymmetric economics; it does not automatically become a durable advantage.
03 · Probe · external
Residential Vantage Network
1/5 signal4/5 potential upside2/5 copy resistance
Turn a consented subset of Mysterium nodes into a last-mile measurement and geo-compliance network—selling observations instead of anonymous egress.
Real home-network vantage could reveal service availability, DNS, latency, loss and geo-policy behavior that datacenter-only probes miss.
Node software, residential presence and settlement may compress the cost of a pilot. Current coverage, stability, consent and ASN diversity are not verified.
The attractive version is recurring B2B measurement with a tiny pilot. The unattractive version rebuilds the network before a buyer appears.
Strongest attackThe unused asset may be imaginary.
Nodes built for traffic relay are not automatically reliable probes. RIPE Atlas already has more than 12,000 probes. Catchpoint advertises 127 last-mile agents within 3,109 agents across 108 countries. ThousandEyes and Kentik also sell distributed measurement. Mysterium may have weaker control, unstable supply, no permission model and no buyer reason to switch.
Cheapest prove / kill
Eight to ten buyer interviews across CDN, platform reliability, geo-compliance and observability. In parallel, five to ten explicitly consented nodes measuring scheduled DNS, HTTP, latency, loss and geo-policy observations.
Kill condition
No two credible design partners name a costly unsolved problem; no stable consented nodes by country/ASN; or required integrity demands a new network rather than a thin layer.
Cost / next decision
24–40 Kairos hours, excluding node instrumentation. 1 week: capability and consent perimeter. 2 weeks: sponsor-approved discovery plus a tiny reliability result. No product build.
Expand evidence, competitors and limits
Mysterium ecosystemPublic residential network description; current node counts remain contested.
RIPE Atlas12,000+ probes and user-defined network measurements.
CatchpointCommercial last-mile, backbone, wireless and cloud-agent incumbency.
Cisco ThousandEyesEndpoint measurement from real user environments.
KentikGlobal synthetic agents plus private deployments.
05 · Not on the podium
Held and rejected
A rejection log is evidence that selection happened. It is not a graveyard of unloved ideas.
Agent-payable access
Interesting settlement fit; no current MN demand signal; incumbents can expose APIs and alternative payment rails.
Provenance / proof-of-origin
Attestation and consent are not established. Build and trust burden arrive before demand.
Grant-funded censorship resistance
Mission fit is real; funding is capped and the likely transport build is heavy.
VPN retention stack
Promising once churn, payment, cancellation, cohort and support evidence can be joined.
Poland incentive economics
Valuable causal audit; presently a capital-allocation test rather than a standalone opportunity.
Org-wide intelligence layer
A means, not a September opportunity. Material pursuit remains gated separately.
06 · The honest promise
Three to five tested. Only survivors recommended.
Four weeks can produce a decision portfolio. It cannot guarantee that evidence will cooperate.
CommitSeriously test 3–5 candidates
Across internal and external lanes, using one comparable structure, explicit attacks, kill conditions and a rejection log.
Do not fake3–5 equally strong signals
If only one candidate reaches recommendation-grade evidence, the pack should say one.
Delivery dependencyData and owner access
Blocked time is logged and surfaced early. It is not hidden inside analysis volume or converted into a judgment about a person.
Signal quality beats symmetry and count.
Authority boundary. This page authorizes no material build, launch, outreach programme, transfer or spend. Applicable Opportunity Schedule gates remain in force before material pursuit.
07 · Provenance
What this ranking rests on
Company reports, public primary sources, explicit inference—and named unknowns.
01
Mysterium All-Hands Analysis · 08-07
02July proxy result, paid-test share, Payments Orchestration, Adyen plan and torpedo contracts.
Kairos Master State
03Current constraints, corrected claims, hypothesis board and authority gates.
RIPE Atlas · official
04Global probe network, customised measurements and current probe scale.
Catchpoint · official
05Current commercial agent inventory and last-mile positioning.
Cisco ThousandEyes · official
06Endpoint agents and measurements from user environments.
Kentik · official
Global synthetic agents and private deployments.
08 · Machine record
Same verdict. Structured.
A compact reading layer for collaborator agents; no hidden evidence and no added authority.
- Verdict
- One lead, one internal challenger, one external probe. They are not three equal recommendations.
- Lead
- Paid-Test Conversion Strike: strongest current signal, limited durable moat.
- Challenger
- Payment-Recovery Torpedo: clean near-cash causal test, structurally copyable.
- External probe
- Residential Vantage Network: strongest asset inversion, no Mysterium demand evidence and real incumbency.
- Scope
- Test three to five. Recommend only candidates that clear the evidence threshold.
- Authority
- No material pursuit, build, outreach programme, transfer or spend is authorized by this page.