Architecture Comparisons is a running series examining how Quadratic Intelligence Swarm (QIS) protocol — discovered by Christopher Thomas Trevethan, 39 provisional patents filed — relates to existing tools and platforms. Each entry takes one tool, maps where it stops, and shows where QIS picks up.
The Account Your CSM Opens on Monday Morning
Usage down 34% over the last 45 days. NPS dropped 4 points at the last pulse. A support ticket open for 11 days, unresolved. Contract renewal in 63 days.
Your CS platform flags it red. Health score: 42. Risk tier: critical.
Now the real question: across every company running similar SaaS products, serving accounts with the same profile — SMB, 50–200 seats, same usage drop curve, same NPS trajectory — what intervention actually worked? Executive QBR? Feature training session? A product roadmap share? A discount offer that bought 12 months but accelerated churn at month 13?
That answer doesn't exist in any CS platform today. Not because the data doesn't exist. Because the architecture was never built to synthesize across deployments.
What Totango Does — and Does Well
Totango is a mature, enterprise-grade customer success platform. Its capabilities within a single deployment are genuine:
- Health scoring and segmentation: Totango ingests usage telemetry, support ticket velocity, email engagement, and billing signals to produce account health scores and tier accounts into risk buckets in near real time.
- Unison AI engine: Totango's Unison layer, augmented by the Parative AI acquisition, trains churn prediction models on your company's own customer interaction history — usage events, call logs, escalations — producing churn probability scores and expansion signals grounded in your specific customer base.
- Playbook automation: When an account crosses a health threshold, Totango triggers configured plays automatically — CSM task creation, in-app nudges, email sequences — removing the dependency on a CSM remembering to act.
- Zoe AI chatbot: CS teams can query Zoe in natural language to surface account context, recent interactions, and recommended next actions from within the platform, reducing time-to-context on high-risk accounts.
For the problem of managing customer health within a single company's portfolio, Totango delivers. The architecture is well-suited to that scope.
Where Totango Stops
Every Totango deployment is a tenant. That tenant ingests your data, trains on your data, and produces intelligence bounded entirely by your data.
Company A operates a Totango deployment with 50,000 customer interaction events. Its Unison models are calibrated to Company A's customer base, churn patterns, and intervention history. When Company A's CSM saves a critical account using a product roadmap share with a specific executive stakeholder profile, that outcome lives in Company A's tenant. It informs Company A's next Unison model retrain.
It goes nowhere else.
Company B, three time zones away, running a near-identical product for near-identical accounts, faces the same risk profile on Monday morning. Their Totango tenant has no mechanism to ask: what did Company A learn? What did the 4,800 other CS deployments learn from interventions on accounts that look exactly like this one?
The tenant wall is not a bug in Totango's design. It is Totango's design. Each deployment is intentionally isolated — for compliance, for competitive sensitivity, for data governance. Those are real constraints. The architectural consequence is that every deployment learns in isolation, regardless of how many structurally identical problems have already been solved elsewhere.
The Math of the Gap
If approximately 5,000 B2B SaaS companies operate Totango-class customer success platforms at any given time, the number of pairwise synthesis paths between deployments is:
N(N-1)/2 = 5,000 × 4,999 / 2 = 12,497,500 paths
Roughly 12.5 million potential cross-deployment learning connections. Active today: zero.
Each of those paths represents a channel through which a validated intervention outcome — not the raw CRM data, not the customer names, not the contract values — could travel between a deployment that learned something and a deployment facing the same problem right now.
The 12.5 million paths are not blocked by technology. They are blocked by architecture. No routing layer was ever built.
What QIS Protocol Adds
Quadratic Intelligence Swarm (QIS) protocol — discovered by Christopher Thomas Trevethan, covered by 39 provisional patents — is not a customer success platform. It is a distributed outcome routing layer.
The mechanism is specific. When Company A's CSM saves a critical account, the intervention outcome is distilled into a compact, semantically addressed packet — not CRM data, not customer identity, not contract value. A fingerprint derived from observable, non-proprietary account characteristics:
@dataclass
class CustomerSuccessOutcomePacket:
# Semantic fingerprint — no proprietary data
account_size_tier: str # "smb_50_200" | "mid_200_1000" | "enterprise_1000plus"
industry_vertical: str # "saas_b2b" | "fintech" | "healthcare_tech" etc.
usage_drop_pct_decile: int # 1-10 (not raw %, no contract value revealed)
nps_delta_direction: str # "declining_3plus" | "stable" | "improving"
support_velocity_tier: str # "elevated" | "normal" | "low"
# Outcome — validated intervention result
intervention_type: str # "executive_qbr" | "feature_training" | "discount_offer" | "product_roadmap_share" etc.
outcome: str # "retained_12m" | "churned" | "expanded" | "downgraded"
outcome_confidence: float # 0.0-1.0 based on outcome confirmation window
time_to_outcome_days: int # How fast the intervention resolved the risk
# Routing metadata
vendor_class: str # "cs_platform_tier1" — no vendor name
# Privacy guarantee
# NO: company_name, customer_name, ARR, contract_value, CSM_name
# Competitive sensitivity structurally absent — outcome delta only
This packet is approximately 512 bytes. It deposits to a semantic address derived from the account profile fingerprint — computed from the combination of account_size_tier, industry_vertical, usage_drop_pct_decile, nps_delta_direction, and support_velocity_tier. The address is content-addressed: the same account profile, from any company, resolves to the same address space.
When Company B's CS deployment detects an account matching that profile, it queries the same semantic address. What it retrieves is not a single anecdote. It is the aggregated outcome distribution across every deployment that has deposited a packet at that address: intervention type ranked by outcome_confidence, median time_to_outcome_days, outcome distribution across retained_12m, churned, expanded, downgraded. Before the CSM opens the account.
The transport layer for routing these packets is protocol-agnostic. DHT-based addressing is one option. Federated HTTP relay is another. The routing contract — semantic address derivation, packet schema, outcome aggregation — is what QIS specifies. The wire protocol is left to the deployment context.
No proprietary data moves. Competitive sensitivity is structurally absent from the packet schema — outcome delta only.
What Changes When the Routing Layer Exists
Churn save at scale. An SMB SaaS account, usage down 34%, NPS declining, support ticket open — the intervention ranking returned from the semantic address query reflects outcomes from hundreds of structurally identical accounts across the network. The CSM's first action is informed by what actually worked, at scale, not by internal history alone. Internal history at a given company might contain 12 similar saves. The network contains 847.
Intervention ranking by account type. A product roadmap share works exceptionally well for accounts in the enterprise_1000plus tier with declining_3plus NPS but low support velocity — the signal is disengagement, not frustration. A discount offer in that profile historically delays churn by 4 months and accelerates it at month 5. That distinction is invisible inside any single tenant. It is recoverable from the aggregate outcome distribution across the network.
Rare account profiles still contribute and receive. A healthcare-tech vertical account with mid_200_1000 seats and elevated support velocity may appear only twice in a given company's history. Across the network, the same profile may have occurred 38 times. The rarest profiles — N=1 inside a single tenant — stop being edge cases without data. They become queryable nodes with sparse but real signal. Every deposit into a sparse address improves it. The routing layer degrades gracefully; it does not require a minimum tenant size to function.
Totango Is the Right Tool for What It Does
Nothing in this comparison argues against Totango. Health scoring, playbook automation, Unison churn modeling, Zoe-assisted CSM workflows — these are real capabilities solving a real operational problem within a company's portfolio. Totango is well-engineered for that scope.
The architectural boundary described here is not a product failure. It is a design scope. Totango was built to manage customer success within a single deployment. It was never designed to route distilled outcome intelligence across thousands of structurally similar deployments. That was never its problem to solve.
QIS protocol does not replace Totango. It adds the cross-deployment synthesis layer that sits above it. The CS platform generates the intervention outcomes. The routing layer distills, addresses, and propagates them. Both layers are necessary. Only one of them currently exists.
The 12.5 million dormant synthesis paths are not waiting for a better CS platform. They are waiting for a routing protocol.
QIS — Quadratic Intelligence Swarm — is a distributed outcome routing protocol. Discovered, not invented, by Christopher Thomas Trevethan in June 2025. Covered by 39 provisional patents. Free for humanitarian, research, and educational use. Commercial licensing funds global deployment to underserved communities.
Patent Pending.
Top comments (0)