Marketing Data Science Manager

Setting coding standards, evaluating new libraries for the team, understanding the trade-offs between different modelling approaches for business goals, and performing high-level code reviews. You won't be writing production code daily, but you'll understand it deeply.

Making strategic decisions on data warehousing platforms (e.g., Snowflake vs. BigQuery), designing the overall marketing data model and schema, and ensuring efficient data access for your team. You'll understand query optimisation at a high level.

Governing the entire BI ecosystem for Marketing, defining key performance indicators (KPIs), setting dashboarding standards for your team, and presenting high-level insights from these tools to the C-suite. You'll ensure consistent, reliable reporting.

Leading the selection and implementation of new CDPs, defining the enterprise-wide customer data and event taxonomy, and ensuring the CDP effectively supports audience activation and personalisation strategies for the team.

Designing the cloud infrastructure for the entire marketing data science function, managing budgets for cloud resources, and overseeing the MLOps strategy for model deployment and monitoring. You'll manage vendor relationships and ensure scalability.

Partnering closely with Finance to model marketing budget scenarios, forecast ROI for major campaigns, and use these platforms to justify investments and report on the P&L impact of marketing activities. You'll speak their language.

• Director of Marketing Data Science (your line manager)
• Chief Marketing Officer (CMO) and other Marketing Leadership
• Head of Product and Product Managers (for A/B testing and feature impact)
• Finance Director and Finance Business Partners (for budget allocation and ROI reporting)
• Head of Sales and Sales Operations (for lead scoring and customer segmentation)
• IT and Data Engineering teams (for data infrastructure and pipelines)

• Marketing agencies (for campaign performance analysis)
• Technology vendors (CDPs, cloud platforms, analytics tools)
• Industry peers and thought leaders (for best practices and emerging trends)

Customers expect hyper-personalised experiences across all touchpoints, often in real-time. This requires models that can operate at the 'edge' (e.g., on a website, in an app) with very low latency. As a manager, you'll need to strategise how your team builds and deploys these systems, moving beyond batch processing to instantaneous decision-making.

Stream Processing Architectures (e.g., Kafka, Flink) · Low-Latency Model Deployment (e.g., FastAPI, ONNX) · Feature Stores for Real-time Features · Reinforcement Learning for Personalisation

• This quarter: Research existing real-time personalisation platforms and their capabilities.
• Next quarter: Identify one marketing use case where real-time personalisation would have significant impact and scope out a pilot project for your team.
• Month 6: Work with Data Engineering to understand the feasibility of building a low-latency feature store.
• Month 9: Oversee your team's first deployment of a real-time model for a specific marketing touchpoint.

Quick win: Start by integrating a simple real-time recommendation engine (even if rule-based initially) into one of our digital properties to understand the operational challenges.

While A/B testing is great, it's not always feasible or ethical for every marketing intervention. The ability to infer causality from observational (non-experimental) data is becoming critical for understanding true campaign incrementality and making robust strategic decisions, especially in 'walled garden' environments where direct experimentation is limited.

Difference-in-Differences (DiD) & Synthetic Control Methods · Propensity Score Matching (PSM) & Inverse Probability Weighting (IPW) · Instrumental Variables (IV) & Regression Discontinuity Design (RDD) · Directed Acyclic Graphs (DAGs) for Causal Modelling

• This quarter: Read 'Causal Inference for The Brave and True' online to get a solid theoretical foundation.
• Next quarter: Identify a past marketing campaign where A/B testing wasn't possible and challenge your team to apply a causal inference technique (e.g., DiD) to estimate its impact.
• Month 6: Collaborate with an academic or consultant specialising in causal inference to review your team's approach and results.
• Month 9: Present the findings from your team's causal inference work to Marketing leadership, highlighting the practical implications for future campaigns.

Quick win: Start by using DAGs to map out the assumed causal relationships for one of our key marketing KPIs. It's a great way to visualise potential confounders and biases.