You will own our scoring models and analytical work as products — not as projects that get handed over, but as things with users, decisions attached to them, a roadmap, and measurable business outcomes.

Roughly half of your time goes to turning analysis into decisions: framing the questions worth answering, working with the analyst and data scientists to get to real root causes rather than plausible stories, and converting what we learn into concrete product and policy changes. The other half goes to the business: making sure credit, risk, operations, and commercial stakeholders understand what our models do, trust them appropriately, and get what they need from the roadmap.

This is a role for someone who is comfortable saying "that number doesn't mean what you think it means" to a senior stakeholder, and equally comfortable sitting with a data scientist to pressure-test whether a lift in a metric is real.

Required:

• 4+ years in product management, product analytics, business analysis, or a risk/strategy role where you owned outcomes rather than tickets. At least 2 of those years working closely with data or analytics teams.
• Strong analytical foundation: you can write your own SQL, work through a dataset independently, and form a view before anyone briefs you on it.
• Demonstrated experience taking a messy business question through to a root cause and a decision — with an example you can walk us through in detail.
• Working knowledge of how predictive models are built and evaluated. You don't need to build them, but you must be able to hold a real conversation about target definition, sampling, overfitting, and the standard evaluation metrics (AUC/Gini, KS, precision/recall) and what they mean commercially.
• Comfort with experimentation: A/B tests, control groups, statistical significance, and the common ways these go wrong.
• Excellent written and verbal communication with non-technical stakeholders. You can make a threshold decision legible to someone who has never seen a confusion matrix.
• The judgement to push back — on a stakeholder request that isn't worth doing, on an analysis that hasn't found the real cause, on a model being used outside its intended purpose.

Nice to have:

• Background in credit, lending, BNPL, banking, insurance, or fraud — familiarity with concepts like PD, credit limits, roll rates, vintage analysis, collections, or regulatory constraints on automated decisioning.
• Experience with model governance or model risk management.
• Experience defining a customer portfolio or segmentation framework that was actually adopted.
• Python for analysis; familiarity with BI tools (Metabase, Power BI, Tableau, or similar).