Senior / Lead Data Scientist – Media Targeting and Media Mix Optimization

About this role

We are looking for a Senior/Lead Data Scientist with strong expertise in Media Targeting and Media Mix Optimization to design, enhance, and optimize marketing investment strategies using advanced statistical modeling, machine learning, and optimization techniques. The ideal candidate will have experience building scalable optimization solutions that help maximize marketing ROI and improve budget allocation across channels.
This role requires excellent Python programming skills, strong statistical foundations, and the ability to translate complex analytical findings into actionable business recommendations.

Key Responsibilities:

• Design and build customer segmentation models (K-Means, GMM, DBSCAN) on large-scale transaction data to power media targeting strategy.

• Engineer features from raw transaction data — RFM variants, spend trajectories, recency decay — to support segmentation and downstream modeling.

• Validate clusters for statistical robustness and business interpretability, and translate segment-level patterns into clear, actionable narratives using SHAP and similar explainability techniques.

• Calculate and interpret competitive metrics (Spend Index, Wallet Share) to inform targeting and positioning decisions.

• Develop and maintain Bayesian marketing mix models (PyMC/Stan) from first principles, including hierarchical structures and multi-stage/chained architectures with proper uncertainty propagation.

• Build adstock and saturation transformations to model channel-level response curves and extract actionable insights from posterior distributions.

• Design and analyze causal attribution studies (geo experiments, DiD, Synthetic Control) to calibrate and validate model outputs against real-world lift.

• Build constrained and multi-objective optimization models (scipy, CVXPY) to recommend budget allocations across channels, respecting business constraints and floors/ceilings.

• Disaggregate coarse budget plans into monthly/channel-level media plans using temporal disaggregation techniques.

• Integrate Gen AI/LLM tools into analytics workflows to automate narrative generation, insight summarization, and reporting.

• Partner with marketing, media, and business stakeholders to translate analytical outputs into clear recommendations and decision-support tools.

• Document methodology, assumptions, and model limitations in structured write-ups to ensure reproducibility and transparency across the team.

• Work independently against a defined brief, proactively flagging risks, data gaps, or blockers to stakeholders.

Segmentation & Media Targeting

• Customer segmentation: K-Means, GMM, DBSCAN — understands the underlying mathematics, not just the API

• Cluster validation: silhouette score, stability testing, business interpretability

• Feature engineering on transaction data: multi-window RFM, spend trajectory, recency decay, time spine construction

• SHAP explainability: interpreting feature importance and translating it into plain-English segment narratives

• Competitive metrics: Spend Index (issuer) and Wallet Share (merchant) — calculation and correct interpretation, including network coverage limitations

• Temporal disaggregation: Denton-Cholette or equivalent — distributing coarse budgets into monthly media plans

• Media budget allocation logic: heuristic channel splits and MMO response curve integration

• Data filtering: BIN/ICA logic for issuer data, merchant_parent_name for merchant data

Bayesian Modelling & Optimisation

• Bayesian regression: model specification from scratch (likelihood, priors, hierarchy) — not just library calls

• PyMC and/or Stan: model building, MCMC sampling, convergence diagnostics (R-hat, ESS, divergences)

• Hierarchical Bayesian modelling: partial pooling, multi-level structures, handling sparse group data

• Chained / multi-stage modelling: sequential model architectures with correct uncertainty propagation (Monte Carlo through the chain, not point estimates)

• Constrained nonlinear optimisation: scipy.optimize, CVXPY — budget allocation, channel floors/ceilings, portfolio constraints

• Multi-objective optimisation: Pareto frontier generation, weighted utility functions, conflicting objective handling

• Causal attribution: geo experiment design and analysis, Difference-in-Differences, Synthetic Control, experiment-to-model calibration

• Discontinuous / partial regression: piecewise regression, change-point detection (PELT, BOCPD, Bayesian), regression discontinuity design

• Adstock and saturation transformations: geometric, Weibull, Hill function — parameter specification via priors, response curve extraction from posteriors

• ArviZ for Bayesian diagnostics and posterior visualisation

• Bachelor's or Master's degree in Statistics, Mathematics, Computer Science, Data Science, Economics, Operations Research, Engineering, or a related quantitative field.

• 5+ years of hands-on experience in Data Science, Marketing Analytics, Media/Audience Targeting, or Marketing Mix Optimization.

• Strong programming experience in Python, with solid SQL skills for data extraction and analysis at scale.

• Excellent grounding in regression modeling, applied statistics, machine learning, feature engineering, and model validation.

• Experience working with large marketing, sales, or transaction-level datasets, including customer segmentation and audience targeting.

• Experience developing optimization or decision-support models for budget allocation, scenario planning, or media mix decisions.

• Hands-on experience with LLM APIs (OpenAI, Anthropic/Claude) for building analytics-adjacent workflows — narrative generation, summarization, or automated insight write-ups.

• Prompt engineering for structured outputs (e.g., generating segment personas, JSON-formatted summaries, or reproducible analysis narratives).

• Experience integrating LLMs into data pipelines — e.g., calling APIs programmatically from Python, parsing/validating…