Three revenue engines compound on the same ledger. Per-seat SaaS funds the wedge: operators adopt for ops pain. Lender data subscriptions fund the orchestration: capital partners read attested vintage from the vault. BPS on capital flow funds the endgame: every dollar that lands on an address pays a small toll for the substrate underneath. Each engine pays for the next. $10M ARR is reachable on the first two; the third is what makes the multiple data-infra, not vertical-SaaS.
▸ Operators pay SaaS · lenders pay for data + inference · capital flow pays BPS · operators never pay a markup on their capital cost · Allometry is never the broker. The substrate stays neutral; the recursion stays clean.
Each engine has a different buyer, a different pricing model, and a different time-to-revenue. They share one thing: the vault gets thicker every time any of them fires. The operator using SaaS thickens the vault for the lender; the lender subscribing thickens the substrate for the next operator; the BPS on flow funds the next lender integration. The recursion is the moat — and the moat compounds on every transaction.
Operators adopt the 16 modules to fix urgent ops pain. The signed evidence accumulates underneath.
Capital partners read attested vintage from the vault — continuous QoE, on-demand Pulse inference, the same evidence a credit committee underwrites against.
Every dollar of capital deployed through the substrate pays a small basis-point toll. Always charged to the lender, never the operator. Baked into the lender's margin, not stacked on top of the borrower's cost.
The reason this isn't just "three revenue lines stapled together" is that each engine produces the conditions for the next. Operator SaaS revenue funds the engineering depth that makes the lender data feed credible. Lender subscriptions fund the partner integrations that make capital flow programmatic. Flow BPS funds the institutional-grade attestation that enables the next tier of capital. The recursion is structural, not coincidental.
Operator subscription revenue funds the engineering depth — hash-chaining at write-time, schema normalization across 16 modules, Pulse Engine training — that makes the vault credible as a lender-facing data feed. Without operators paying for the platform, the data layer never gets built.
Lender subscriptions fund the next tier of integration: ZKP attestation production-grade, covenant resolution APIs, smart-contract verifier circuits, on-chain anchoring. The flow rails are the infrastructure that turns "verifiable data" into "automated capital deployment."
The basis points on capital flow fund direct integrations with operator software, free pilots for design partners, AUM-based incentives for early operators. Capital lands faster on existing operators, which makes Allometry adoption a no-brainer for the next cohort. The recursion closes.
The recursion is the difference between "vertical SaaS with a roadmap" and "infrastructure that compounds." A pure-SaaS company has to win the SaaS market and then somehow earn the right to do data infra. Each engine here unlocks the funding for the next — and the moat thickens at every layer.
Modeled across three years from pre-seed close. The mix is the point: at $10M ARR, ~30–40% of revenue is already data-infra-shaped (lender subscriptions + early flow BPS). Same total ARR, different category. The mix is what changes the multiple. See §04 for what that means.
| Engine | Year 1 · Pre-seed → Seed | Year 2 · Seed → Series A | Year 3 · Series A → Growth |
|---|---|---|---|
| Engine 01 · Operator SaaS | 5–15 ops × $50k = $250k–$750k | 30–80 ops × $50–75k = $1.5–6M | 100–200 ops × $75–100k = $7.5–20M |
| Engine 02 · Lender subscription | 1–2 design partners · free | 5–15 × $100k = $500k–1.5M | 25–50 × $150–300k = $3.75–15M |
| Engine 03 · BPS on flow | $0 | $100k–500k (small) | $500k–2M (scaling) |
| Total ARR (mid-band) | ~$500k | ~$3.5M | ~$14M |
| Data-infra share | 0% | ~15% | ~30–35% |
▸ Conservative band. Aggressive band trends $20–35M Year 3 with the right cohort scaling. ▸ Engine 02 ARR is gated by operator coverage density — lenders sign when there's a critical mass to underwrite a portfolio against. That density crosses around 25 vintaged operators per industry vertical. ▸ Engine 03 ARR is gated by ZKP production-readiness and the first signed covenant agreement; we model it as small Year 2, compounding Year 3+.
Pure vertical SaaS at $10M ARR exits at 5–8× — call it $50–80M. The same $10M ARR with 30–40% data-infra contribution exits at data-infrastructure multiples — 10–20× — call it $150–300M. The recursion is the literal reason for the difference. Investors pricing the round care less about the dollars than about which curve those dollars sit on.
Comparables (vertical operator SaaS):
What the multiple says: the buyer is paying for retained operator revenue. Growth is linear in operator count.
Comparables (data infrastructure):
What the multiple says: the buyer is paying for a substrate that compounds with every API call and every basis point of flow. Growth is super-linear in coverage × usage × flow.
Each engine has its own pricing logic. Operators are priced by size + module count; lenders are priced by tier × access depth + per-inference usage; flow BPS is priced by capital deployed, baked into the lender's margin.
| Operator tier | Annual ARR | Includes |
|---|---|---|
| SMB · <$10M revenue | $25k–$50k | 4 modules · 1 loop · standard support |
| Mid-market · $10–100M | $50k–$150k | 8–16 modules · multi-loop · dedicated implementation |
| Enterprise · $100M+ | $150k–$500k+ | 16 modules · multi-entity · custom covenant mapping |
| Lender tier | Annual ARR | Includes |
|---|---|---|
| T1 · Banks · ABL · regional | $50k–$150k | Read access to <100 vaults · monthly QoE refresh · basic Pulse inference |
| T2 · Infrastructure debt · mid-tier | $150k–$300k | Cohort access · streaming QoE · full Pulse inference API · covenant resolution |
| T3 · Insurance-linked · institutional | $300k–$500k+ | Full attestation · on-chain anchoring · co-developed covenant types · audit-grade access |
| T4 · Pension · rated paper | $500k+ + custom | Methodology disclosure · Big-4 co-attestation · multi-vintage cohort access |
▸ Plus per-inference usage: $0.50–$5 per Pulse inference call. Scales with the lender's underwriting workflow, not Allometry's operator count.
| Capital tier | BPS on flow | Mechanism |
|---|---|---|
| T1 · Working capital | 5–10 bps | Drawn-line BPS, baked into the lender's spread |
| T2 · Infrastructure debt | 10–20 bps | Origination + lifetime BPS, baked into facility pricing |
| T3 · Insurance-linked | 15–25 bps | Parametric trigger + payout BPS, baked into premium |
| T4 · Rated paper | 10–20 bps | Issuance + servicing BPS, baked into spread to investors |
▸ Charged to the lender / insurer / capital provider — never marked up on the operator's cost of capital. Allometry stays neutral between sides.
The substrate works because both sides — operators and capital — can trust that Allometry isn't taking a hidden cut. Four things we explicitly don't do. Naming them is the architecture.
Both are legitimate. Both are answered structurally, in the architecture of the business model — not by hand-waving away the risk.
▸ Q1Building a credible underwriting model takes multi-year arcs. How does revenue compound before the underwriting model is marketable?
It compounds from Engine 01 and the easiest band of Engine 02, neither of which require a "fully marketable underwriting model" to clear. Operator SaaS revenue starts in month 1; it doesn't depend on lenders existing at all. The first lender subscription (T1 — banks / ABL / working capital) underwrites against six months of vault vintage, not five years — because it's lending against working capital and contracted A/R, not a residual-value model on a 10-year asset. The deeper underwriting models (T3 insurance-linked, T4 rated paper) are real multi-year arcs — and they're where Engine 03 (BPS on flow) becomes huge — but the path to $10M ARR doesn't cross them. It runs on Engine 01 + the T1 band of Engine 02.
Put differently: we don't need to be a credit-rating agency to hit $10M ARR. We need to be a SaaS company that has earned the right to sell a lender-facing data feed. The first lender pilot signed at T1 unlocks $1–5M of Engine 02 revenue without touching the harder underwriting tiers at all.
▸ Q2Capital appetite drifts in and out of asset classes for reasons that have nothing to do with ease of underwriting. How is that not a structural risk?
It is a structural risk for any company tied to a single asset class. The architectural answer is that the substrate is asset-class-agnostic: the vault doesn't know if it's storing data from an EV charging operator, a distributed solar developer, a fiber-build crew, an HVAC service business, or a humanoid-labor fleet. The address-level operating graph is the same primitive. The covenant projection — which fields a lender underwrites against — is tier-specific, not asset-class-specific.
This means we can follow capital appetite across asset classes. If insurance-linked capital pulls back from EV-charging operators in 2027 but rotates into distributed energy storage, the same vault primitive serves both. The cohort migrates; the substrate persists. Diversification is a property of the architecture, not a strategy to be executed.
The downside case is real: if all physical-infrastructure asset classes simultaneously lose capital appetite, Engine 03 (BPS on flow) gets compressed. But Engines 01 and 02 keep paying — operators still need ops software, and lenders that *are* deploying still need attested data. The recursion slows; it doesn't stop.
Engine 01 is producing revenue today. Engine 02 is in design-partner pilot with one T1 lender. Engine 03 is a roadmap line — and we don't claim it as ARR until the first signed flow-BPS agreement is executed. The same honesty band that lives on /the-vault applies here.
The model is unusual — three engines, one ledger, lender-side monetization on top of an operator-side wedge. Below are the questions we get most.
Three reasons. First, regulatory: charging operators a fee tied to capital they receive would make Allometry a loan broker in most jurisdictions, with all the licensing, disclosure, and capital requirements that implies. Second, alignment: any operator-side fee tied to capital introduces the perception that we're routing to specific lenders for our benefit — which kills the wedge. Third, economics: lenders bake the cost of acquisition into their margin; operators pay every dollar out of pocket. The same $1 hurts the operator 5× more than it hurts the lender.
Operators pay for the SaaS that closes their ops loops. The capital that lands as a byproduct is unmarked-up. The substrate stays neutral.
The data-infrastructure model is the clean path. Lenders pay for API access to operator-consented data, plus per-inference compute. That's the same regulatory shape as Plaid, MX, Codat, Modern Treasury — data infrastructure, not loan brokerage. The operator grants consent; the lender consumes data; Allometry never executes the loan, takes possession of funds, or matches a specific borrower to a specific lender on a deal-by-deal basis.
BPS on flow is a different structure — closer to embedded-finance arrangements like Plaid Income or Stripe Treasury. Where it crosses into regulated territory (specifically: when a covenant verifier triggers a smart-contract draw), we operate through partner-licensed entities. We will never represent ourselves as something we aren't.
Marketplaces match borrowers to lenders on a deal-by-deal basis. They take a finder's fee. The marketplace itself doesn't have a proprietary data layer; it's a routing engine. Allometry is the inverse: we sell continuous data infrastructure to lenders that they use to underwrite their own deals, not deals we route to them. The lender's underwriting decision happens entirely on their side; we provide the substrate.
The closest comparable in shape is Plaid for a different vertical. Plaid sells consumer-bank-account data infrastructure to fintech apps; those apps decide whether to extend credit. Allometry sells operator-vault data infrastructure to lenders; those lenders decide whether to extend capital. The vertical differs (consumer fintech vs. physical revenue); the architectural pattern is the same.
Lender data subscriptions price against operator coverage density — a lender won't pay $150k/year to access data on 3 operators. The first paid Engine 02 contract gates on roughly 25 vintaged operators in the lender's target vertical. Until then, the design-partner lender gets free access in exchange for cooperation on covenant structure, methodology validation, and case-study disclosure. That cooperation is worth more than a year's revenue at this stage — it's what unlocks the next 10 lender conversations.
The honest framing: Engine 02 is structurally pilot-stage, not "we couldn't close the contract." It will go paid in Year 2 when coverage crosses density.
Two gates. First, production-grade ZKP attestation must clear sub-100ms reliably on real inference workloads — currently in benchmark, not production. Second, a signed flow-BPS agreement requires a counterparty willing to structure their margin around our spread — which requires a multi-year track record with that counterparty at Engine 02 first. The realistic timeline is Year 2 for small Engine 03 ARR, Year 3+ for material Engine 03 ARR.
We model Engine 03 conservatively in the $10M ARR plan precisely because the multi-year-build risk is real. If it lands faster, the plan accelerates; if it doesn't, $10M is still achievable on the first two engines.
A vertical SaaS company adding a capital product owns both sides of every deal — it's the merchant of record, the originator, or the principal-on-balance-sheet. Toast Capital, Shopify Capital, Square Loans all do this. The advantages are huge customer-LTV and lock-in. The disadvantages are massive capital requirements, credit-risk-on-balance-sheet, and the operator's perception that the SaaS is "marking up" their capital.
Allometry is infrastructure, not principal. We don't originate, don't take credit risk, don't sit on the balance sheet of the deal. We sell the data and proof layer that lets other capital providers compete for the operator's business. The trade-off: less captive LTV per operator, but a substrate that scales without balance-sheet constraints and earns multiple revenue lines per operator over their lifetime. Different shape, different multiple.
The BPS structure is anchored to the cost the lender would otherwise pay to source, diligence, and underwrite the deal. Today that cost is real: regional banks spend $5–15k per ABL underwriting cycle, 4–8 weeks, with non-trivial false-positive rates. The vault collapses that to seconds with cryptographic attestation. The 5–10 bps we charge for that on a $1M working-capital line is $500–1000 — a fraction of what the lender saves on cycle time and analyst hours. The BPS is priced against the lender's existing cost stack, not against the operator's cost of capital. We're cheaper than the alternative for the lender, free for the operator, and net-positive for the deal.
The risk is that lenders eventually want to internalize the substrate themselves. Possible — but every lender that internalizes pays the build-cost; only Allometry has the cross-lender coverage that makes the substrate genuinely composable. The two-sided network protects the pricing.
Modeled assumptions: pre-seed $250k SAFE at $3M cap (8% dilution); seed $5M priced at $30M (~14% dilution + 10% option pool refresh); Series A $20M at $100M post-money (~17% dilution). Founder retains ~45–50% at $10M ARR; ESOP at ~15%; pre-seed angels ~6%; seed lead ~10%; A lead ~13%. Standard infra-grade cap table. Founder control intact through Series A.
If Engine 03 lands faster than modeled and Series A prices higher (say $60M instead of $30M based on data-infra multiples), dilution falls accordingly and founder retention crosses 55%. This is the most material reason to optimize for the mix, not just the ARR — same dollars, different cap table.
▸ The business model rests on the substrate. These are how it's built.
16 operator modules deposit hash-chained evidence into one ledger. Time-in-vault unlocks the lender tiers Engine 02 and Engine 03 are priced against.
/the-vault →The translation layer — every signed atom reshaped into the 16 fields a lender underwrites against. New: Quality of Earnings across the four lender tiers.
/covenant-structure →Signed atoms, ZKP receipts, MCP-callable proofs. The technical surface that makes Engine 02 lender API consumption composable.
/attribution →The four-tier lender ladder Engine 02 and Engine 03 price against — banks, infrastructure debt, insurance-linked, rated paper.
/lending →Operator SaaS funds the wedge. Lender data subscriptions fund the orchestration. BPS on capital flow funds the endgame. $10M ARR is reachable on the first two. The third makes the multiple data-infra, not vertical-SaaS.
Talk to us about capital partnership →