Product
Structure the evidence. Then hand it to a human.
Hola Credit converts a borrower-authorised bank statement into structured cash-flow evidence, bounded by confidence levels and produced by versioned deterministic code. Same inputs, same policy version, same result. Every time.
How it works
Four steps, fully auditable.
Record consent
Before any processing begins, the applicant explicitly authorises the organisation to use their bank statement for a defined credit assessment. No preselected or bundled consent is permitted.
The consent artefact records the purpose, data categories, organisation, retention scope, expiry, and the actor who captured it. Consent is versioned, auditable, and withdrawal is recorded without rewriting history.
Upload securely
The loan officer uploads the bank statement inside the organisation's tenant. PDF is preferred because it preserves layout and text encoding, which makes extraction more reliable. PNG and JPEG are accepted with OCR as a fallback.
The file moves through a defined lifecycle: quarantine, scanning, acceptance or rejection, parsing, then a retention hold with a deletion due date. Statements from FNB Namibia, Bank Windhoek, Standard Bank Namibia and Nedbank Namibia are supported at launch.
Extract and check
Transactions are extracted and categorised, each with a confidence score and source-page provenance. Low-confidence rows are flagged for human review rather than guessed silently. Corrections are recorded as review events with before-and-after values.
This is the only place AI touches the pipeline. AI extracts, it does not decide.
Review the assessment
The assessment summarises the cash-flow evidence across its components, each with its own confidence level, plus a plain-language explanation a lender can read aloud to a colleague.
The lender adds their own notes, applies their own policy, and makes the decision. The decision record is owned by the lender's organisation and stored separately from the assessment.
Rhythm explorer
Feel the rhythm of real income.
Salary underwriting asks one question: does the payslip exist? Cash-flow underwriting asks a better one: what does the money actually do? Drag through an illustrative year and watch the signals take shape.
Illustrative example
Nangula, freelance designer, Windhoek
12 months of net business income, as read from her bank statement
Months read
9 of 12
Income in Sep
N$ 8,200
Month 9 of the statement
Income floor so far
N$ 7,300
Rarely falls below this
Consistency so far
89%
Months inside a steady band
Swing so far
37%
Quietest to busiest month
Illustrative example for demonstration only, not real customer data and not a lending recommendation.
The assessment
Four signals, in the words a lender uses.
Each signal answers one question, and each carries its own confidence level based on the quality and sufficiency of the underlying data.
Income floor
What can this applicant genuinely rely on?
The minimum reliable monthly income the statement can support, derived from consistent credit transactions and discounted for concentration risk and volatility. When data is insufficient, the income floor is null rather than a misleading number.
Consistency
How regular is the rhythm?
How much of the income lands in a steady band around the median. Months inside the band, months outside it, and no reward for one lucky deposit.
Volatility
How wide does income swing?
The spread between the strongest and weakest months, relative to the median. Context for any repayment plan a human might build.
Red flags
What needs a human eye?
Patterns that deserve attention before a decision: gambling-shaped activity, sudden reversals, unexplained round figures. Flags are surfaced, never acted on automatically.
AI extracts. AI does not decide.
The scoring engine does not use a machine-learning model to calculate the assessment. It uses versioned code with fixed rules. Given the same extracted transactions and the same policy version, the engine will always produce the same score. No randomness, no hidden state, no model discretion.
Every assessment records the policy version that produced it, so results are traceable and comparable over time. And because the first score cannot yet be validated against repayment outcomes, it is presented as supplementary evidence, not a prediction.The honest detail is in the FAQ.
See it read a real statement.
The demo walks the full pipeline with a sample file. No account needed.