BuildFi / Alternative Data Engine

Alternative Data Engine.

Six signal streams · one project-and-counterparty score · target sub-six-second decisions. Built from the data sources banks structurally can't see — designed for the Sub-Payment Facility.

UNDERWRITING · ARCHITECTUREPRE-LAUNCHTARGETS BELOWSHARIA-COMPATIBLE
Underwriting · BUILT FROM SIGNAL

Credit scored from the GC and the project — not the SME balance sheet.

Banks underwrite the SME. That's why construction subcontractors don't get credit: they look thinly-capitalised by design, with negative working capital at the start of every job and a cheque register that can't be modelled.

BuildFi underwrites the general contractor and the project. A signed master subcontract, a certified developer escrow, a 24-month main-contractor litigation history, a certified IPC, and 90 days of SME cash-flow velocity — these predict GC-side payment far better than two years of audited financials.

The Alternative Data Engine fuses six signal streams into a tier-A/B/C decision plus advance ratio, holding-fee tier and concentration impact. Underwriting is GC-first, project-second, sub-third.

Signal 01 · Developer escrowDLD / RERA live account status · funding ratio · release cadence
Signal 02 · GC dispute historyADGM Courts / ADJD / Dubai Courts dispute registry · 24-month look-back
Signal 03 · GC payment disciplineLate-payment events > 90 days · IPC certification cadence
Signal 04 · SME cash-flow velocityOpen Banking via Lean / Tarabut · 90-day rolling
Signal 05 · Project-health graphInter-subcontractor exposure · concentration view
Signal 06 · PDC registerSub-side security-cheque volume + post-2022 civil dishonour history
ArchitectureRule-based gates → ML scoring → manual override (CRO sign-off)
OutputTier (A/B/C) · advance ratio (75–90%) · pricing tier · concentration check
RefreshEvery advance · plus weekly portfolio sweep · monthly model review
Anatomy of a decision

Six signals. One tier.

Input · ESCROWDLD · RERA
Input · COURTSADGM · ADJD · DXB
Input · OPEN BANKINGLean · Tarabut
Input · CERTIFICATIONIPC + GC protocol
feature engineering · derived features
Layer · GC HEALTHDisputes × payment cadence
Layer · PROJECT HEALTHDeveloper × escrow × age
Layer · SUB SIGNALCash velocity × PDC × graph
tier rules · concentration caps · Sharia gate · CRO override
OUTPUT · DECISIONTier A · Advance $200K (80% of $250K certified IPC) · 3.0% disc + 2.0%/30d hold · 90 days

Illustrative output only. Actual decisions are made on a per-deal basis, subject to FSRA RegLab admission and Sharia Advisory Board pre-clearance. No credit is being offered at this time.

Model targets · pre-launch

Targets we will hold ourselves to.

These are pre-launch design targets, not realised performance. The platform has not yet originated a single advance. All targets will be refreshed quarterly against actual portfolio data once live, and reviewed at every credit committee.

DECISION LATENCY · TARGET

< 6 sec P95

From API call to tier decision. Auto-approval on repeat SMEs in < 1 sec cached. Final decision within 24–72 hours including manual review and KYC steps.

↳ SLA TARGET · 99.5 %
DEFAULT TARGET · BLENDED

< 2.5 %

Net loss after PDC and civil-recovery, blended across the loan book through cycle. Phase 1 sub-target < 1.5 % on the first $5–10M deployed.

↳ MEASURED QUARTERLY
CONCENTRATION CAPS

20 / 30 / 5 %

Single GC ≤ 20% of book (RegLab: 15%); single developer ≤ 30% (RegLab: 25%); single sub ≤ 5%. Hard rules in the credit engine; not Excel-after-the-fact.

↳ ENFORCED AT ADVANCE-TIME

See the full risk memo. Professional Clients only.

Confidential. Indicative only. Not an offer or financial promotion. Restricted to Professional Clients (ADGM FSRA definition) or equivalent accredited investors. Capital is at risk. Forward-looking statements are estimates and not guarantees.