Castel — Credit Analysis and Scoring Technology

The credit decisioning platform
for African lenders.

Score, instruct, and approve SMB loans against the data that actually describes the borrower — not the data that happens to be the easiest to find. From application intake to the analyst's final decision, on a single platform.

app.helix.mansa-rt.com / applications
Applications
live
Files this month
1,284
↑ 8.4%
Approval rate
61.3%
↑ 2.1 pp
Median latency
1.4s
P95 · 3.8s
Override rate
4.1%
all logged
Reference
Applicant
Amount
Score
Decision
MNS-04-A8F3
Sté Adwoa Textiles
XOF 4,250,000
0.87
Approved
MNS-04-B7E2
Atelier Konaté
XOF 1,180,000
0.92
Approved
MNS-04-C9F4
Boutique M. Diallo
XOF 720,000
0.64
Review
MNS-04-D2A1
Coop. Agricole Thiès
XOF 8,400,000
0.81
Approved
MNS-04-E8B7
Atelier Yaméogo
XOF 380,000
0.38
Declined
MNS-04-F4D9
Restaurant Chez Awa
XOF 2,100,000
0.76
Approved
Designed with banks and microfinance institutions in West Africa
BCEAO
RCCM
AWS
MTN
01 — What's inside

Three things make Castel defensible to your risk committee.

Feature 01

Every score, fully attributed.

Castel never returns a score without a feature attribution alongside it. Risk officers see exactly which signals contributed, in which direction, with what weight. No black boxes. The model can be defended to a regulator, line by line.

Score 0.87 — Approved
MNS-04-A8F3
mobile_money_consistency
+0.34
supplier_payment_pattern
+0.21
bank_statement_volatility
+0.12
geographic_cluster_risk
−0.08
cooperative_membership
+0.18
repayment_history
+0.10
Feature 02

Every decision auditable, end to end.

From the moment a dossier is submitted to the moment a risk officer makes the call, every step is logged with a timestamp, a source, and a hash. Regulators receive the trail on request, in a format they already know how to read.

Decision audit trail
MNS-04-A8F3
+0.00sINGESTapplication.received
+0.18sFETCHbank_statement_24mo · ok (412 transactions)
+0.41sFETCHmobile_money_orange · ok (1,847 events)
+0.62sFETCHsupplier_invoices · ok (38 records)
+0.89sNORMALIZEtimeline.merged · 2,297 events
+1.14sSCOREmodel: tailored · 87 features
+1.42sDECIDEscore: 0.87 · class: pre-approve
+1.43sLOGaudit.write { regulator_facing: true }
+11mREVIEWanalyst · @adwoa.sarpong · approved
Feature 03

Custom models trained on your data.

Starting from your own SMB customer data, we build scoring models adapted to your credit activity. Evolving with your volumes, integrated automatically into your decision processes — not a pre-trained model dropped on top of your stack.

Decision pipeline
01
Application intake
Public portal, documents, KYC
02
Ingestion
Bank · mobile money · BCEAO · RCCM
03
Scoring
Tailored model + feature attribution
04
Analyst decision
Approve · Decline · Override (with reason)

See Castel run on your loan book.

A focused walkthrough with your risk team, on a few representative dossiers from your loan book. We'd like to understand how you decide today, and show how Castel could fit into that.