Clear answers on data, reliability and governed use.

LuminoUnderwrite makes application and renewal review faster and more consistent without removing underwriter judgment. LuminoEvidence complements your data science team by industrialising signal engineering and evaluation. It returns a governed library of engineered signals, validation results for each target outcome, measured incremental lift beyond the current baseline, and a reproducible experiment record showing which signal sets, model specifications and parameters produced the strongest reliable result.

Categorisation identifies what a transaction was. LuminoUnderwrite evaluates what the observed pattern means for an SME credit assessment. It creates behavioural measures across ability, stability, conduct, trajectory, concentration and confidence, then connects them to source evidence and underwriting context.

It does not infer legal structure from bank transactions alone. LuminoUnderwrite extracts declared entity, applicant, owner and guarantor information from the loan package, then reconciles it with bank account-holder evidence and authoritative business-registry or KYB sources. It can distinguish and connect company, sole-trader, partnership, trust and trustee arrangements, while flagging unresolved name, identifier, ownership, address or account-control inconsistencies for human review.

The definitions can stay consistent while calibration changes by operating pattern. Daily takings, invoice-led, wholesale and retainer businesses are measured against the rhythm and seasonality that make sense for them. LuminoEvidence then validates each signal inside the segments where it will be used.

Decisioning platforms execute rules and models in production. LuminoUnderwrite prepares the case evidence and structured assessment those systems need. LuminoEvidence determines which engineered signals add value for a specific outcome, quantifies their incremental lift and preserves the experiment and governance evidence. Approved signals and specifications can then feed your existing model, scorecard, decision engine or workflow.

No. LuminoUnderwrite creates a consistent, evidence-backed assessment for review. LuminoEvidence provides engineered signal candidates and the evidence showing where they add value. Your team decides whether and how to use them in a scorecard, model, policy or strategy, and retains ownership of approval authority, overrides, governance and final production sign-off.

You still receive a fixed-cost answer and the full analysis. The evidence may show that your current approach already captures the signal, the proposed representation is not stable, or the target needs refinement. Knowing what not to build prevents larger production waste.

LuminoEvidence requires representative historical inputs, the decision or treatment that occurred and the downstream outcome. Depending on the use case, that may include applications, behavioural metrics, approvals and declines, exposure, repayment, delinquency, default, loss, recovery, pricing, offer acceptance, renewal, retention or collection outcomes. Not every use case requires every field; the target, observation window, baseline and holdout design are agreed first.

It can support default and loss modelling, scorecard development, automated-approval and referral strategies, decision-engine optimisation, risk-based pricing, pre-approved offers, propensity and next-best-action modelling, early collections, portfolio monitoring and portfolio optimisation. The exact output depends on the available history, outcome quality, decision context and governance requirements.

No. LuminoUnderwrite is designed for SME lending and does not depend on a universal consumer score. It can use business application data, authorised bank transactions, verification results, repayment history and prior relationship data. The data scope, purpose, minimisation, de-identification and permitted use are agreed before analysis.

No. A pilot uses a one-off historical extract, typically account, repayment and transaction history joined to outcomes. Integration is considered only after the evidence supports production use.

Depending on the workflow and permissions, inputs can include application fields, business documents, bank transactions, verification results, repayment history, prior applications, current exposure and policy outcomes. Luminnovate uses only the fields needed for the agreed assessment.

A synthetic extract can be used to prove the pipeline, metric definitions and workflow while privacy and security review proceeds. The measured uplift that informs an investment decision must still come from representative historical data.

Client-specific validated signals and evidence produced from your data are governed by the engagement terms. Luminnovate retains its underlying platform, reusable metric specifications and evaluation methodology. Your data is not used to improve another client's models.

Coverage and account scope are reconciled before behaviour is measured. Missing periods, overlaps, transfers between owned accounts and unobserved activity are made explicit. Where the available history cannot support a reliable signal, the output is withheld or carries a stated coverage and confidence limitation.

One sponsor who owns the decision, one data contact who can produce the extract and a workshop to define the target and success bar. The aim is hours of client involvement, not a parallel internal project team.

Each production candidate can carry a written specification, version, lineage, target definition, evaluation design, holdout result and commercial effect. The evidence is designed to support model inventory, validation and change-control processes.

LuminoUnderwrite prioritises interpretable behavioural metrics and conclusions that connect back to observed source evidence. LuminoEvidence documents why a metric earned a place in the decision and where its evidence does or does not generalise.

Reliability is assessed against the intended decision, not declared universally. Evidence can include data coverage, held-out performance, incremental value beyond the current baseline, time and cohort stability, sensitivity to reasonable definition changes and ongoing drift monitoring. A signal that does not clear the agreed bar is rejected or limited to further review.

The pilot is fixed scope and fixed price, agreed before work starts. Production licensing is annual and scales with the size and use of the portfolio. Commercial options can be discussed once the decision, evidence bar and available data are understood.

The pilot is designed to answer whether the evidence supports production use before live integration begins. A production path is then defined around your existing stack, controls, monitoring and workflow. Integration can be API or batch and is designed to add intelligence without forcing a platform replacement.