Senior Credit Risk Analyst (PJN) at Weaver Fintech Ltd

Purpose of the role:

• This role exists to turn fragmented data into profitable lending decisions. The business originates credit across two very different products — a high-volume, thin-margin, short-tenor BNPL book and a longer-dated retail instalment loan book — and the data that should inform those decisions currently sits in silos: bureau feeds in one system, transactional and behavioural data in another, merchant and collections data elsewhere again.

The Senior Credit Risk & Decision Scientist is accountable for three connected outcomes:

• Data optimisation across silos — building a single, governed, decision-ready view of the customer by integrating bureau, internal and alternative data sources, and eliminating duplicated or unused data spend.
• Scorecard build and monitoring — developing, validating, deploying and continuously monitoring application, behavioural, affordability and collections scorecards across both product lines.
• Account origination — designing and optimising the end-to-end origination decision engine: policy rules, cut-offs, limit assignment, affordability assessment, fraud screening and champion/challenger strategy.
• Success is measured commercially. The role is expected to demonstrably improve risk-adjusted return — higher approval rates at constant or lower loss, lower cost of data acquisition, and faster decisioning — not simply to produce models.

Key Responsibilities:

Data Optimization across business silo's:

• Map the full credit data estate across origination, servicing, collections, merchant/partner and marketing systems; document lineage, ownership, refresh cadence, quality and cost of every source.
• Design and own a consolidated credit data mart / feature store that serves modelling, decisioning and reporting from a single set of definitions, removing conflicting versions of the same metric.
• Integrate and rationalise the three data families the business depends on:
• Bureau data — scores, enquiry data, tradeline and payment-profile history, adverse and judgment data, affordability indicators.
• Internal data — application, transactional, repayment and arrears behaviour, customer tenure, cross-product holdings, merchant and basket-level data, servicing and contact history.
• Alternative data — bank transaction / open-banking data, device and digital footprint, telco and payment-behaviour signals, geospatial and psychometric indicators where lawful and predictive.
• Run a continuous cost-to-value assessment of every paid external data source: quantify incremental lift per Rand of bureau or alternative-data spend and retire, renegotiate or re-sequence calls that do not pay for themselves.
• Implement tiered / cascaded data-call strategies so expensive attributes are only purchased where they change the decision.
• Define and enforce data quality standards, reconciliation controls and monitoring for all decision-critical feeds; own the resolution path when a feed degrades or a bureau attribute drifts.
• Partner with Data Engineering to productionise pipelines, and with Product and Finance so that a single agreed set of risk and profitability metrics is used across the business.
• Scorecard Development:
• Own the full model lifecycle for application, behavioural, affordability, fraud-propensity and collections scorecards across BNPL and retail loan portfolios.
• Construct reliable modelling samples: target definition, performance and outcome windows, exclusions, reject inference, and correction for the short performance windows and rapid repeat-usage cycles typical of BNPL.
• Engineer and select features from bureau, internal and alternative sources; apply appropriate binning, WoE/IV analysis, segmentation and multicollinearity treatment.
• Build models using the right tool for the job — logistic regression and scorecard scaling where explainability and regulatory defensibility are required; gradient boosting and other machine-learning techniques where lift justifies them, supported by explainability output (SHAP or equivalent).
• Validate rigorously: out-of-time and out-of-sample testing, Gini/KS/AUC, calibration, stability, segment-level performance and fairness/disparate-impact testing.
• Produce model documentation to internal model-risk and audit standard, and present models to the Credit or Model Risk Committee for approval if needed.
• Support deployment into the decision engine and sign off implementation testing — confirming that the scored production population matches the development specification, attribute for attribute.

Scorecard monitoring and model governance:

• Own the full model lifecycle for application, behavioural, affordability, fraud-propensity and collections scorecards across BNPL and retail loan portfolios.
• Construct reliable modelling samples: target definition, performance and outcome windows, exclusions, reject inference, and correction for the short performance windows and rapid repeat-usage cycles typical of BNPL.
• Engineer and select features from bureau, internal and alternative sources; apply appropriate binning, WoE/IV analysis, segmentation and multicollinearity treatment.
• Build models using the right tool for the job — logistic regression and scorecard scaling where explainability and regulatory defensibility are required; gradient boosting and other machine-learning techniques where lift justifies them, supported by explainability output (SHAP or equivalent).
• Validate rigorously: out-of-time and out-of-sample testing, Gini/KS/AUC, calibration, stability, segment-level performance and fairness/disparate-impact testing.
• Produce model documentation to internal model-risk and audit standard, and present models to the Credit or Model Risk Committee for approval if needed.
• Support deployment into the decision engine and sign off implementation testing — confirming that the scored production population matches the development specification, attribute for attribute.

Account Origination and Decision Strategy:

• Own the origination decision logic end-to-end: pre-screen, identity and fraud checks, bureau and alternative data calls, scorecard execution, affordability assessment, policy rules, cut-off setting, limit assignment and referral treatment.
• Optimise approval-rate versus loss trade-offs using swap-set analysis; quantify the profit impact of every proposed cut-off or policy change before it goes live.
• Design initial credit limit and limit-increase strategies for BNPL, balancing basket-size conversion and merchant experience against exposure at risk.
• Build and run a disciplined champion/challenger and A/B testing programme, including test design, sample sizing, holdout maintenance and readout, so that strategy changes are evidence-based rather than opinion-based.
• Reduce friction and decision latency: increase straight-through-processing rates, cut manual referrals, and shorten time-to-decision without increasing loss.
• Work with Fraud to separate first-party and third-party fraud losses from credit losses, and ensure origination strategy treats them differently.
• Feed origination insight back into Product, Merchant and Marketing teams — which channels, merchants, baskets and segments originate profitable accounts, and which do not.

Commercial and Profitability impact:

• Translate analytics into commercial recommendations with quantified rand-value impact, and defend them to Credit, Finance and Exco.
• Contribute to credit loss forecasting and to the annual budget and business planning cycle.

Other:

• Act as the analytical authority in Credit Committee forums; present clearly to non-technical audiences.
• Mentor junior analysts, review their code and models, and raise the analytical standard of the team.