The Business Intelligence (BI) team at d.light owns the full data and analytics stack — from raw data ingestion and transformation, through our data warehouse, to the dashboards, automations and tools that power business strategy & decisions across the company. The team is built around two pillars: Data Platform, which builds and runs the warehouse, pipelines, automation and AI stack, and Analytics Delivery, which partners with departments to produce the tools, visibility and reporting the business runs on.
The Decision Analytics Lead sits somewhere between the two. We have a long and growing queue of business questions that a dashboard alone cannot answer — what is likely to happen next? which customers to act on? Which lever actually moves the number? and whether an intervention worked. This role exists to work through that queue, initially alongside the Director of BI, on problems spanning credit and collections, commercial, supply chain and finance. This role is judged on decisions supported and business impact, not on the sophistication of the method.
What the Role Entails
- Own the hard questions: Take ambiguous, high-stakes questions from leadership and departments, sharpen them into something answerable, and see them through to a defensible answer and a decision.
- Forecast and predict: Build forecasts and predictive models that are accurate enough to plan against and transparent enough to be trusted.
- Score, segment and cluster: Produce risk scores, segmentations and prioritised action lists that operational teams can act on directly.
- Design and evaluate experiments: Set up tests and control groups for business interventions, then give an honest read-out of what moved and what did not.
- Work across the modern data stack: Get your own data and build your own models — SQL and dbt in Redshift one day, a Python script or Jupyter notebook the next, Tableau when a visual is the right way to land the point.
- Get into the business: Spend real time with commercial, credit, supply chain, finance and country teams, including in the field, so your analysis reflects how we actually operate.
- Communicate and land the decision: Turn complex work into clear recommendations for non-technical audiences up to senior leadership. Be honest about assumptions and uncertainty without being paralyzed by them.
- Make your work reusable: The datasets, metrics and definitions you design should not live only with you. Work with Data Platform to build the good ones into our warehouse models so the wider team can use them, and with Analytics Delivery to absorb recurring outputs.
What Success Looks Like
- The backlog moves. Questions that have sat unanswered for months get credible, documented answers.
- Decisions change. Leaders and department heads make different, better calls because of your analysis.
- Your work reaches operations. Scores, forecasts and prioritised lists are in live use, not sitting in a deck.
- Your best ideas become shared assets. A new metric, dataset or way of looking at the business that you design gets built into our data models and used by other analysts and teams.
- Business teams seek you out with their messiest questions, and your analysis is documented well enough to hand over or rerun.
Requirements
- A business-first analyst, not a back-room modeller. You are as comfortable in a commercial or credit review as you are in a notebook, and you would rather answer a real operational question well than build an elegant model nobody uses.
- 5+ years answering complex analytical questions in an operational business, or a convincing demonstration of that depth by another route. What matters is a track record of analysis that changed a decision.
- Strong SQL and the ability to build your own data models against a cloud data warehouse — you can find and stitch together the data you need yourself (Redshift and dbt experience a plus).
- Strong Python and Jupyter skills for analysis, automation and modelling, with the discipline to write code others can follow.
- Working knowledge of applied statistics, with exposure to machine learning techniques such as regression, classification, forecasting and clustering, applied pragmatically.
- Able to scope an ambiguous question into a plan: what would answer this, what data exists, what is good enough, and when to stop.
- Comfortable with imperfect operational data — you investigate outliers and data quality problems rather than quietly modelling around them.
- Excellent communication and documentation skills, including explaining method, assumptions and uncertainty to a non-technical audience. Familiarity with Tableau or similar is expected.
- Exposure to consumer credit, PAYGo, lending, distribution or supply chain is a strong plus, as is any experience where analysis had to survive contact with an operational team.
- A passion for the work d.light does and the customers we serve, with enthusiasm for an open, startup environment.
Benefits
- Competitive Remuneration Package
- Medical Cover
- Pension
