WOAM Capital · Investor Appendix A
Confidential · March 2026

The Compound
Risk Architecture

How two independent risk dimensions are held simultaneously by AI to produce investment decisions no human underwriting system could make
"The absence of formal financial records does not indicate the absence of trustworthiness. It indicates the absence of a system capable of reading the trust signals that actually exist. AI changes that — and adds a structural protection layer that makes trust signals unnecessary for a significant portion of the portfolio."
Contents of this appendix
1The structural problem with trust-only underwriting
2The two independent risk axes
3The investment matrix and what lives in each quadrant
4The substitution principle — how trust and assets trade off
5Dynamic investment structuring — same pipeline, different outputs
6The chain quality filter
7Compound monitoring — same event, different responses
8Why this is a durable competitive moat
What no human underwriter
can do at scale,
AI holds as standard practice
I
Trust and asset coverage are independent. A portfolio company can have strong behavioural signals and no collateral. Or minimal history and a clear, productive, title-retainable asset. The compound model serves both — and every combination in between.
II
Scores update continuously. A portfolio company who enters in the low-trust quadrant and makes six consecutive on-time payments has demonstrated trustworthiness through behaviour. Her score moves. Her chain release terms improve. The model learns.
III
Each investment is structured differently. The pipeline output is not "approved" or "declined." It is a recommended capital structure — specific mix of grant, asset-backed working capital, and cash working capital — calibrated to the candidate's exact position in the two-dimensional risk space.
IV
The moat is not the algorithm. The algorithm is reproducible. The moat is the combination of portfolio history, community trust relationships, and platform adoption that makes every component of the compound model more accurate over time — and that no new entrant can purchase.
WOAM Capital · Confidential · March 2026
Section 1

The structural problem with
trust-only underwriting

Every existing model for nano-business investment in emerging markets collapses to a single question: is this person trustworthy? The question is correct. The mechanism for answering it is not.

01
The NGO intermediary model

A programme officer who knows the candidate personally vouches for her. The institution trusts the officer's judgment. Capital flows through the vouching relationship. Trust is borrowed rather than assessed.

Bottleneck: activation speed of institutional relationships. Growth rate is bounded by how many programme officers can be onboarded. Portfolio scale follows NGO activation, not market demand.
02
The credit history model

Repayment records, bank statements, and formal income evidence are used to predict future behaviour. The assumption is that past institutional behaviour predicts future institutional behaviour.

Bottleneck: 1.7 billion people globally have no formal financial history. The model excludes by design the population it claims to serve. The data does not exist for the candidates who need capital most.
03
The group lending model

Grameen-style peer guarantee. A small group co-guarantees each other's repayments. Social accountability enforces debt service. No individual collateral required.

Bottleneck: designed for debt, not equity. Revenue share is contingent on income — there is nothing to enforce when income genuinely falls. Group guarantee applies to fixed obligations, not income-contingent ones.
04
The collateral model

Physical assets are pledged against the investment. If the obligation is not met, the lender recovers the asset. The collateral is the trust mechanism.

Bottleneck: the population with no credit history also has no formal collateral. Assets exist informally but are not legally encumberable. The model requires formalisation before it can be used — creating a catch-22 that WOAM Capital's formalisation grant is designed to break.

"Each of these models solves one dimension of the trust problem. None of them simultaneously assess behavioural trust signals and provide structural protection. The compound model does both — and AI is the technology that makes holding both simultaneously feasible at scale."

Section 2

Two independent axes.
Continuously scored.

The compound model scores every candidate on two dimensions that are genuinely independent. A high score on one does not predict a high score on the other. Both are required to determine what investment structure is appropriate.

Axis 1 · Behavioural

Trust Score

Continuous assessment of behavioural evidence for financial reliability

  • Mobile money transaction history. 6–12 months of inflow patterns. Regularity of business-type counterparties. Absence of large irregular outflows. Seasonality consistent with stated franchise type.
  • Social graph density. Number of distinct regular payment counterparties. Community embeddedness score derived from payment network analysis — without any individual disclosing their connections explicitly.
  • Peer circle validation. 3 nominated community members independently interviewed by the AI pipeline. Consistency of their accounts against the candidate's declaration. Narrative coherence across interviews.
  • Video assessment consistency. Monthly check-in sessions analysed for response latency, narrative consistency month-on-month, vocal stress patterns, and cross-referencing of stated income against payment data.
  • Post-investment behaviour (ongoing). Revenue share payment cadence, platform engagement, formalisation progress velocity, peer network activity. Score updates every month — it is never fixed.
What a high trust score means: the pipeline has strong multi-source evidence that this portfolio company will honour the revenue share relationship. It does not mean she will definitely pay — it means the behavioural evidence is consistent and reinforcing.
Axis 2 · Structural

Asset Coverage Score

Proportion and quality of working capital deployed as title-retained productive assets

  • Coverage ratio. What percentage of the working capital component is deployed as a specific, identifiable, title-retained physical asset rather than as cash? 100% asset coverage means zero cash working capital disbursed.
  • Asset-income fit score. How directly does this specific asset generate the income from which revenue share is paid? A smartphone for a savings group coordinator — high fit. The same phone for a manual repair technician — low fit. Scored 0–1 per asset-franchise combination.
  • Asset recoverability. Can WOAM Capital practically recover or deactivate this asset if the revenue share relationship is abandoned? Smartphones (MDM profile) and motorcycles (vehicle registration) score high. Solar systems (installed infrastructure) score medium. Cash scores zero.
  • Asset depreciation curve. Does the asset retain value over the revenue share relationship period? Motorcycles and solar panels depreciate slowly. Smartphones depreciate faster. The collateral value schedule affects the effective protection over the break-even window.
  • Title clarity. Can WOAM Capital's ownership position be clearly recorded and enforced in this geography? Vehicle registry (Rwanda, Kenya) — high. Serial number + MDM profile (phones) — medium. Informal marking (tools) — lower. Legally encumbered collateral scores highest.
What a high asset coverage score means: the structural incentive for the portfolio company to complete the revenue share relationship is high — because the asset is making her money every day, and she cannot own it outright until break-even is reached.
Section 3

The investment matrix.
Four quadrants. One pipeline.

Every candidate lands somewhere in this space. The pipeline's job is not to produce a binary approve/decline. It is to identify where in the matrix the candidate sits — and to design an investment structure appropriate to that position.

Trust Score (Axis 1) → High
Low Asset Coverage
High Asset Coverage →
II
High trust · Low asset

Trust-led investment. Standard structure.

The savings group coordinator. The community organiser. The educator whose business model is relationship-based, not equipment-based. No natural asset to title-retain. Approved on behavioural evidence alone.

Approved — standard

Revenue share collection through platform rails critical here — no asset backstop. Chain release requires minimum trust score maintained at trigger point.

I
High trust · High asset

The ideal investment. Both mechanisms active.

Strong behavioural evidence reinforced by structural protection. The compound model operates at full intensity. Lowest failure rate. Fastest break-even. Highest quality chain anchor — her referrals receive expedited pipeline assessment.

Approved — priority

Eligible for largest investment ticket. Best candidate for chain seeding. Both social incentive and structural incentive operating simultaneously.

IV
Low trust · Low asset

Not investable at this stage.

Neither behavioural confidence nor structural protection is available. The pipeline cannot make a defensible investment decision. This is not a permanent rejection — it is a pathway question.

Not yet

Recommended pathway: 90 days of mobile money activity, or identification of a productive asset use case. Re-application triggers expedited reassessment.

III
Low trust · High asset

The unlocked quadrant. Asset-led investment.

The most excluded population. No mobile money history. No NGO affiliation. Invisible to institutional data systems — but with a clear, specific, productive use for a title-retainable asset. Under a trust-only model: declined. Here: approved with structure.

Approved — structured

100% working capital as title-retained asset. Zero cash component. Enhanced monitoring cadence. Revenue share rate may be set at upper end of range to compensate for lower trust confidence.

Low Trust ↓
Section 4

The substitution principle.
How trust and assets trade off.

The compound model is not two separate thresholds applied independently. It is a continuous substitution function: additional asset coverage can compensate for a trust deficit, and additional trust evidence can compensate for low or zero asset coverage. The pipeline holds this continuously.

A candidate with a trust score of 65 — below the 75 auto-approve threshold in the trust-only model — and 80% asset coverage may produce a compound approval score above the threshold. A candidate with a trust score of 90 and zero asset coverage is approved on behavioural evidence alone. The substitution curve describes how much of one compensates for how much of the other.

0%
Asset
25%
Asset
50%
Asset
75%
Asset
100%
Asset
Trust
85+
Approve
Approve
Approve
Approve
Approve
Trust
70–84
Review
Approve
Approve
Approve
Approve
Trust
55–69
Decline
Review
Approve*
Approve*
Approve*
Trust
<55
Decline
Decline
Decline
Review
Review
* Approved with structured terms: 100% working capital as asset, enhanced monitoring cadence, revenue share rate at upper range. "Review" = human analyst assessment triggered.
"The substitution function is not arbitrary. It is initially set from the theoretical risk model and calibrated continuously against actual portfolio performance — the pipeline learns which combinations of trust score and asset coverage actually predict revenue share completion, and the curves move accordingly. By Month 18, the model is more accurate than its initial specification."
Section 5

Dynamic investment structuring.
Same pipeline. Four different outputs.

The pipeline does not apply uniform terms to all approved candidates. It designs a capital structure for each one. These four cases illustrate how the compound model produces different investment recommendations for candidates who would all pass a simple trust threshold — or be declined by it entirely.

1
Savings Group Coordinator
Claudine U.
Musanze, Rwanda
Q2 · High trust
Trust score
88
  • 24 months mobile money history — 22 regular inbound counterparties consistent with stated group membership
  • Peer circle: 3 of 3 accounts consistent with her declaration
  • Video assessment: 4 sessions, high narrative consistency
  • Cooperative registry: applied, pending — formalisation in progress
Asset coverage score
18
  • Business model is relationship-based — no high-fit asset identified beyond smartphone
  • Smartphone recommended: $140 device, MDM profile, IMEI registered to WOAM Capital
  • 18% asset coverage = $78 of $432 working capital in title-retained asset
  • Remaining $354 as cash working capital for group management expenses
Pipeline recommendation
Approved — standard structure. $468 formalisation grant (unconditional) + $140 smartphone (title retained, MDM profile) + $292 cash working capital. Revenue share 10% of total income. Break-even estimated Month 23 at standard income trajectory. Platform collection at source from Month 3 when cooperative bank account confirmed. Chain release trigger: $432 cumulative, minimum trust score 80 maintained.
2
Repair Collective Lead
Asel D.
Bishkek region, Kyrgyzstan
Q1 · Ideal
Trust score
82
  • 18 months mobile money history — consistent service payment inflows
  • KOMPANION MFI data: 14-month repayment record, no missed payments
  • Peer circle: 3 of 3 consistent; one peer independently mentions municipal contract potential
  • Video: 3 sessions, strong consistency, income figures match payment data within 8%
Asset coverage score
91
  • Tool kit: $380 of specialist repair equipment. Serial numbers registered. Asset-income fit 0.94 — tools are the direct income generator
  • Smartphone: $140. MDM profile. Asset-income fit 0.71 for job management and customer payments
  • 91% asset coverage = $393 of $432 working capital in title-retained assets
  • Remaining $39 as cash float for consumables
Pipeline recommendation
Approved — priority structure. $468 formalisation grant + $380 tool kit (WOAM Capital title, serial registered) + $140 smartphone (MDM profile) + $39 cash float. Revenue share 10% of total income. Estimated income uplift: 3× in 18 months post-formalisation (municipal contract access). Break-even estimated Month 12. Chain release eligible at Month 13 — fastest possible release, both mechanisms active. Priority candidate for chain seeding.
3
Community Kitchen
Fatima N.
Kigali outskirts, Rwanda
Q3 · Asset-led
Trust score
51
  • No mobile money history — operates entirely in cash, no smartphone
  • No NGO affiliation, no prior institutional relationship
  • Peer circle: 2 of 3 consistent; one peer unavailable for interview
  • No video history — first intake session only. Insufficient baseline for consistency scoring
Asset coverage score
96
  • Solar refrigerator: $380. Direct income generator — food preservation enables school supply contract she cannot currently fulfil. Asset-income fit 0.96
  • Smartphone: $52 entry device. MDM profile. Enables mobile money transition
  • 96% coverage: $414 of $432 working capital in title-retained assets
  • Trust-only model: declined. Compound model: approved with structure
Pipeline recommendation
Approved — structured, asset-led. $468 formalisation grant + $380 solar refrigerator (WOAM Capital title, serial registered, NGO field officer delivery confirmation) + $52 smartphone (MDM profile) + $0 cash. Revenue share 10% of total income. Enhanced monitoring: fortnightly video check-in for first 6 months. Asset condition photo upload monthly. Chain release requires minimum trust score 65 at break-even — she must build trust history to unlock her chain. This is by design: the asset protects the investment; the monitoring builds the trust score.
4
Market Trader
Zara K.
Nairobi, Kenya
Q1/Q2 · Mid-range
Trust score
71
  • 9 months mobile money history — income variable and seasonal, consistent with market trader profile
  • Peer circle: 2 of 3 consistent. One peer interview shows income estimate 40% higher than declared — flagged for discussion, not disqualification
  • No MFI relationship. Market registration exists but informal
  • Video: 2 sessions. Adequate consistency. One stress response on income question — noted
Asset coverage score
58
  • Cargo bicycle: $220. Replaces daily 3-hour walk to market. Enables 2 additional trading days per week. Asset-income fit 0.88
  • Smartphone: $80. MDM profile. Mobile money transition critical for platform collection
  • 58% coverage: $250 of $432 in title-retained assets
  • Remaining $182 as cash working capital for stock. Higher cash component given income variability
Pipeline recommendation
Approved — blended structure with milestone disbursement. $468 formalisation grant + $220 cargo bicycle (vehicle registration, WOAM Capital title) + $80 smartphone (MDM) disbursed immediately. Cash working capital ($182) disbursed at Month 3 after first 3 revenue share payments confirmed — milestone tranche reduces exposure during highest uncertainty window. Income discrepancy between declaration and peer estimate flagged for Month 2 video check-in clarification. Revenue share 10% total income. Break-even estimated Month 25 at conservative income.
Section 6

The chain quality filter.
Not every break-even triggers a chain.

The investment chain releases three referrals when a portfolio company reaches her break-even threshold. But in the compound model, the trigger has two conditions: the financial threshold must be met, and the portfolio company's current trust score must be at or above a minimum. Both must be true simultaneously.

This prevents a specific failure mode: a portfolio company who limped to break-even — multiple missed payments, recovered through asset-backed collection, deteriorating community relationships — releasing a chain of three referrals under her endorsement with a compromised referring relationship. The chain quality filter ensures that every chain anchor is a portfolio company who has genuinely demonstrated reliability, not merely one who happened to reach the financial threshold.

01
Break-even reached
Cumulative revenue share ≥ $432 (working capital recovered)
Financial condition

The pipeline confirms cumulative revenue share receipts have reached the break-even threshold. This is calculated from platform payment records — both automatic deductions (platform collection at source) and manual remittances. Milestone is timestamped and immutable.

What this confirms

WOAM Capital has recovered its working capital component. The formalisation grant remains sunk — that is by design and is not recovered here. The equity stake in the cooperative is now effectively free to WOAM Capital. The financial condition is necessary but not sufficient.

↓ And simultaneously ↓
02
Trust score check
Current trust score ≥ minimum threshold at trigger point
The behavioural condition

At the moment of break-even, the pipeline checks the portfolio company's current trust score — not her intake score, but her score today. This reflects all post-investment behaviour: payment cadence, platform engagement, video consistency, formalisation completion, community network health.

Minimum thresholds by entry quadrant

Q1 entry (high trust, high asset): minimum 75 at trigger. Q2 entry (high trust, low asset): minimum 75. Q3 entry (low trust, high asset): minimum 65 — she must have built trust history post-investment to unlock her chain. This is the mechanism that turns the unlocked quadrant into a full participant in the chain model.

↓ Both true → Chain releases ↓
03
Chain release
Three referrals enter pipeline. Quality weighted by referrer's compound score.
What the chain carries

Each referring portfolio company's compound score at the moment of release determines the weight her endorsement carries in the nominees' pipeline assessment. A Q1 portfolio company releasing a chain at trust score 88 provides a strong endorsement — her nominees are fast-tracked. A Q3 company releasing at minimum trust score 65 provides a standard endorsement — nominees are assessed independently, with her referral noted but not weighted heavily.

The self-selection effect

High-quality portfolio companies release high-quality chains. They know the model, they have demonstrated what it takes to succeed, and their nominees are likely to be similar to them — because social networks are homophilic. Over time, the portfolio self-selects for quality: the most successful companies produce the most productive chains, and the chains grow fastest in the communities with the highest portfolio concentration.

↓ If financial condition met but trust score below threshold ↓
04
Circuit breaker
Trust threshold not met. Chain held. Nominees released via standard pipeline.
What happens

The portfolio company has reached break-even financially but has not demonstrated sufficient reliability to release a chain under her endorsement. Her three nominees — who have been conditionally approved and waiting — are not penalised. They enter the standard pipeline immediately, assessed independently. WOAM Capital absorbs the slightly higher assessment cost.

What this protects

The nominees are protected from waiting indefinitely for a referrer whose reliability has declined. The portfolio quality of the chain is protected — low-trust referrals do not cascade. And the referring portfolio company retains an incentive to rebuild her trust score, because a subsequent 90-day window gives her a second opportunity to release her chain under her own endorsement.

Section 7

Compound monitoring.
Same event. Three different responses.

The compound model's greatest operational value is not at the point of investment. It is in the ongoing monitoring relationship — where the interaction of trust score and asset coverage determines what a missed payment means, and what the appropriate response is.

Three portfolio companies each miss their Month 6 revenue share payment. Under a single-dimension monitoring system, the response is identical for all three — a payment reminder, possibly an escalation protocol. Under the compound model, the same event produces three completely different assessments and three different responses.

Case A · Q1 entry
Asel's repair collective
Trust 82 → Asset 91
82Trust now
91%Asset cvg
M6Payment missed
Event: Month 6 revenue share payment not received. Tool kit confirmed in use — NGO field officer monthly check confirmed active operation last week.
  • Mobile money inflows unchanged from Month 5 — income not declining
  • Platform engagement: 4 modules accessed this month — actively using business tools
  • Video check-in Month 5: high consistency, mentioned cash flow timing issue with new client
  • Asset condition: confirmed operational, no deterioration signals
Likely cash flow timing, not default intent. Grace period 14 days, automated message only. Trust score unchanged. No human escalation. Probability of resolution: very high.
Case B · Q2 entry
Claudine's savings group
Trust 88 → Asset 18
74Trust now
18%Asset cvg
M6Payment missed
Event: Month 6 payment not received. Trust score has declined from 88 at intake to 74 — a 14-point drop over 6 months. Low asset coverage means no structural backstop.
  • Mobile money inflows down 45% vs Month 4 — income genuinely declining
  • Platform engagement: zero for 3 weeks — disengagement signal
  • Video check-in Month 5 was cancelled, rescheduled, then completed 10 days late
  • Peer network: 3 of 22 regular counterparties have stopped transacting — possible group breakup
Business distress likely — not strategic default. Urgent video check-in within 48 hours. NGO programme officer notified. Platform business support module activated for savings group management. Trust score adjusted to 68. Chain release threshold review flagged.
Case C · Q3 entry
Fatima's kitchen
Trust 51 → Asset 96
58Trust now
96%Asset cvg
M6Payment missed
Event: Month 6 payment not received. Low trust entry, now at 58 — below the 65 minimum required for chain release. High asset coverage: solar refrigerator confirmed operational via fortnightly check protocol.
  • Mobile money inflows stable — income not declining. Payment not made but income exists
  • Platform engagement: active — accessed food safety certification module twice this week
  • Video check-in: completed on schedule. No stress signals on income questions. Cited "confusion about payment date"
  • Asset: refrigerator operational, school supply contract confirmed in progress
Income exists; payment not made — possible confusion about mechanism, possible testing behaviour. Asset secure. Immediate platform notification with clear payment instructions and deadline. If not resolved in 7 days: NGO field officer visit to confirm asset status and clarify relationship terms. Trust score maintained at 58 pending resolution. This is precisely the case the asset backstop was designed for.

"A human case manager handling 200 portfolio companies cannot hold three simultaneous variables per company and route each case individually. A compound monitoring system running on the portfolio in real time can — and does, continuously, without human involvement until a human response is specifically warranted."

Section 8

Why this is a
durable competitive moat.

The compound risk architecture is not a feature. It is the product — and its defensibility increases with every investment made, every month of monitoring data collected, and every chain generation completed. Three components of the moat cannot be purchased by a new entrant at any price.

I
The substitution curves are trained on portfolio data that does not exist yet

The initial substitution function — how much asset coverage compensates for a trust deficit — is set from theory. Its calibration comes from actual portfolio performance. By Month 18, with 500+ portfolio companies across multiple geographies and franchise types, the model has been trained on data that no competitor possesses. A new entrant starting today begins with theoretical parameters. WOAM Capital by Month 18 operates with empirically validated ones. The gap widens every month.

II
The community relationships that make the asset model work cannot be built remotely

Title-retained asset delivery requires local sourcing relationships, NGO field officer networks, and community trust that took WOAM fourteen years to build. A competitor can build an AI pipeline in months. It cannot replicate the relationship through which a solar refrigerator is delivered to Fatima in Kigali, confirmed operational, and monitored monthly by someone she knows. That relationship is the asset backstop — and it is not available for purchase.

III
The credit history product is a compound output — not a feature that can be added later

The portable financial identity that WOAM Capital's portfolio companies accumulate — transaction history + formalisation record + video assessment consistency + asset management behaviour + revenue share completion — is a compound data product that requires all dimensions to be collected from Day 1. A competitor who adds behavioural assessment or asset-backing later cannot retroactively build the multi-dimensional history. The credit history product is only possible for portfolio companies who entered the compound model at intake.

Investable quadrants
3 of 4
The compound model serves three quadrants that single-dimension models cannot all serve simultaneously. The trust-only model serves Q1 and Q2. The asset-only model serves Q1 and Q3. Only the compound model serves all three.
Estimated failure rate reduction
~30%
Asset backing reduces effective failure rate from 10% to approximately 7% annually by creating a structural incentive to complete the revenue share relationship that voluntary models cannot match.
Model improvement cadence
Monthly
Substitution curves recalibrate monthly against actual portfolio performance. The model becomes more accurate every 30 days — automatically, without human intervention, from the data the portfolio itself generates.
New entrant replication time
3–5 yr
Estimated minimum time for a well-funded competitor to replicate the combination of portfolio data, community relationships, and asset infrastructure that makes the compound model accurate. The window closes with each month of operation.

The compound model does not make
better decisions than a human underwriter.
It makes decisions that no human underwriter
could make at this scale, in this time, at this cost.

At 200 investment decisions per day, across six geographies, four languages, twelve franchise types, and five asset categories — with each decision requiring continuous two-dimensional risk assessment, dynamic investment structuring, real-time monitoring, and chain quality filtering — there is no human system that approximates what the compound model delivers as a matter of routine operation.

This is not a claim about artificial intelligence in general. It is a specific claim about this specific architecture, applied to this specific problem, with this specific combination of behavioural and structural risk dimensions. The claim is falsifiable. The proof is in the portfolio.

WOAM Capital · Appendix A: The Compound Risk Architecture · Confidential · March 2026