Finance data in Ghana is scarce too. Same forge, new domain.
Mobile Money moves a huge share of everyday transactions in Ghana, but the fraud-signal data behind it never leaves the telcos and banks that hold it — not for students, not for independent fraud- detection research. Finance is Dataset Foundry's second line family, deliberately non-insurance (no frequency/severity model to reuse) to prove the platform pattern holds outside insurance, not just inside it. Mobile Money and Micro-Credit & Susu Scoring are both live today. Ghana is today's default country, not a restriction — every generator takes a country parameter, chosen per dataset rather than fixed by the line.
seed 42 — this exact seal, every time
One platform, one line at a time
Mobile Money and Micro-Credit & Susu Scoring — same domain, two different product shapes — are both live. Structurally realistic data, deterministic and seeded, mimicking the shape of real scarce data rather than generating noise.
Mobile Money
Available nowTransaction-level fraud-signal logs — velocity, SIM-swap recency, wallet tenure, agent cash-out channel — plus a continuous anomaly score alongside the binary fraud label. Toggleable two-stage model adds loss given fraud for flagged transactions, driven by Cash-Out status and agent channel. Deterministic, seeded, ready to forge.
Generate a datasetMicro-Credit & Susu Scoring
Available nowSusu-group contribution and micro-credit repayment data — missed-contribution history, occupation-sector informality, group guarantee presence, and loan-to-savings ratio driving a two-stage default-probability / loss-given-default model. Deterministic, seeded, ready to forge.
Generate a datasetRelationships tested and confirmed in every batch
These aren't hidden in the noise — they're the reason each module exists. Every dataset forges the same directional relationships between its factors and outcome, so your classifier, regression, or classroom exercise has something real to recover. Both lines are two-stage now, each with its own split: Mobile Money runs four hypotheses on fraud probability and two on loss given fraud; Micro-Credit & Susu Scoring splits its four evenly, two on default probability and two on loss given default. Both stage modes are a runtime choice on each line's own generator page.
Mobile Money — fraud & loss risk
Transaction velocity (trailing 24h) Fraud probability
Higher fraud probability
A rapidly rising count of transactions in the trailing 24 hours is the structuring/“smurfing” signal — cycling stolen funds through many small transactions before a wallet gets frozen.
SIM-swap recency Fraud probability
Higher fraud probability
The single largest loading in the model — a SIM swap within the last 24 hours is the strongest individual fraud signal in the Mobile Money literature this line is anchored against.
Wallet tenure Fraud probability
Higher fraud probability (newer wallets)
A wallet opened within the last month has no legitimate transaction history to compare against and is disproportionately the vehicle a fraud ring picks; the loading eases — and briefly goes negative — for the oldest wallets.
Agent cash-out channel Fraud probability
Higher fraud probability (Cash-Out rows only)
Only applies to the ~8% of transactions that are Cash-Outs. An unmanned ATM cash-out carries the largest loading of the four channels; a Bank Partner cash-out sits below the reference band, inheriting the bank's own KYC checks.
Cash-Out status Loss given fraud
Higher loss given fraud
Once cash is disbursed at an agent, it can't be reversed — the point of no return in Mobile Money fraud response. A transaction that stays inside the digital rails is more often freezable and clawable-back before the money moves again.
Agent cash-out channel (recoverability) Loss given fraud
Higher loss given fraud (ATM), lower (Bank Partner)
The same field the frequency-side hypothesis above reads, asking a different question: given a Cash-Out transaction is already fraudulent, how much of it is recovered. Bank Partner Cash-Out inherits the bank's own KYC and reversal machinery; ATM Cash-Out is fully anonymous, with no attendant able to intervene once cash is dispensed.
Micro-Credit & Susu — default risk
Missed-contribution history Default probability
Higher default probability
A member's own record of showing up with their susu contribution is the closest thing an informal lending scheme has to a credit-bureau record.
Occupation-sector informality Default probability
Higher default probability
Informal-sector income is more volatile than formal-sector wage income, a documented driver of microfinance default.
Group guarantee presence Loss given default
Lower loss given default
Fellow group members have a direct stake in recovering part of the shortfall once a default has already happened, improving recovery rather than preventing the default itself.
Loan-to-savings ratio Loss given default
Higher loss given default
A loan sized well beyond what the borrower has accumulated in the susu fund leaves less of the borrower's own money on the table to recover against.
Configurable length, configurable model
Pick a record count, a seed, and a stage mode, and each line's deterministic model — Mobile Money's fraud probability plus a correlated anomaly score, optionally with loss given fraud for flagged transactions, or Micro-Credit & Susu's default probability / loss given default — runs server-side and hands you a CSV, manifest, and citation block. Two-stage is the default on both lines; single-stage is available for a leaner schema. Same seed-and-cite reproducibility as every insurance line.
Both finance lines are live — more domains queued
Mobile Money and Micro-Credit & Susu Scoring are both shipped now,
each with its own hypothesis table and methodology page. What's next
is scoped in the roadmap but not scheduled while the freeze holds: a
new, non-insurance health line family, plus further-out
queued lines (agribusiness, logistics, SME tax).
Other countries' Mobile Money and Micro-Credit markets are on the roadmap too, opt-in per generation — Ghana stays the default, not a ceiling.