Data Scientist

• Benchmarking Methodology: Own the statistical methodology behind customer-facing benchmarking, end to end.

• Comparability Modelling: Build and validate a model that determines which organizations are genuinely comparable, using features such as service mix, pricing, scale, and geography.

• Defensibility Thresholds: Decide where the line sits between a comparison Onit can defend and one it cannot, and design the product's behaviour on both sides of it.

• Data Quality Programme: Profile data quality across multiple source systems and lead a measurable improvement programme against a defined target, partnering with Data Engineering on remediation.

• Entity Resolution: Define the approach for resolving duplicate and conflicting records across systems that were never designed to reconcile with one another.

• External Data Assessment: Assess whether missing attributes can be responsibly supplemented from external sources, and set the evidential bar for doing so.

• Governance & Privacy: Work with Product on methodology and governance for pooled, anonymised cross-customer analytics, including aggregation thresholds.

• Communicating Uncertainty: Explain uncertainty clearly to non-statisticians and help the product communicate confidence honestly to end users.

• 7+ years in applied data science, analytics engineering, or a closely related quantitative role, including ownership of methodology for externally facing analytics.

• Strong applied statistics: sampling, variance, confidence intervals, and sound judgement about small samples (e.g., able to explain why a three-item average may or may not be publishable).

• Practical, production experience with entity resolution, record linkage, deduplication, or master data management.

• Strong SQL, and Python or Ruby for analysis and modelling; hands-on experience working in Databricks (notebooks, Spark/PySpark, Delta tables) or a comparable cloud data platform.

• Experience diagnosing and remediating data quality problems at scale, including reasoning about why data is missing before deciding what to do about it.

• Rigorous validation habits: defines baselines and success metrics before building, and reports honestly against them.

• Ability to communicate uncertainty to non-technical stakeholders without overstating confidence or being unhelpfully vague.

• Working proficiency with AI tools (e.g., Claude, GitHub Copilot) to accelerate analysis, code, and documentation, with the statistical judgment to critically validate AI-generated output.

• Experience building benchmarks, rate comparisons, scoring models, or indices intended for external consumption.

• Experience with anonymisation, aggregation thresholds, or privacy-preserving analytics.

• Experience working with a cloud data warehouse or lakehouse (Databricks preferred).

• Experience in a regulated or commercially sensitive domain, ideally legal.

• Experience defining data quality SLAs or acceptance thresholds with business stakeholders.