As a Marketing Science Analyst, you will play a key role in monitoring, investigating and optimizing marketing mix models (MMM) used by hundreds of brands across major e-commerce and advertising platforms. Working within a data-driven environment, you will analyze production model alerts, identify anomalies and troubleshoot issues affecting the accuracy and reliability of marketing performance insights. You will combine statistical knowledge, analytical judgment and technical skills to distinguish data quality problems from modeling issues and recommend effective solutions. Using Python, SQL and AI-powered coding tools, you will investigate complex analytical workflows and improve model configurations. You will collaborate closely with Data Science and Customer Success teams, translating technical findings into practical business recommendations. This fully remote opportunity offers hands-on exposure to Bayesian modeling, marketing analytics and emerging AI-assisted technologies while contributing directly to the quality of insights used by global brands.
Model monitoring and investigation: Monitor production marketing mix models and investigate alerts involving declining model fit, unusual Return on Ad Spend (ROAS) results and unexpected Customer Acquisition Cost (CAC) fluctuations.
Develop investigation hypotheses, analyze relevant datasets, test potential causes and document findings, conclusions and recommendations.
Identify and distinguish modeling-related issues from data quality problems, including broken data feeds, incorrect campaign labels, reporting delays and other inconsistencies.
Prioritize investigations according to business impact, maintain accurate records and ensure issues are addressed within established timelines.
Marketing mix model optimization: Review and adjust production model configurations, including Bayesian priors, seasonality, holidays, promotions, trends and marketing channel classifications.
Evaluate modeling parameters such as adstock, Hill saturation and other statistical components when investigating performance anomalies.
Compare alternative model configurations against existing production models, assessing statistical fit, channel-level results and overall business plausibility.
Validate proposed configuration changes, promote approved adjustments and document rejected recommendations with supporting analysis.
Maintain comprehensive records of model configurations, modifications, validation results and technical decisions.
AI-assisted data analysis: Use AI-powered coding agents, including Claude Code, to accelerate investigations, analyze model behavior and identify data discrepancies.
Review and interpret Python and SQL code to understand modeling pipelines, investigate analytical workflows and troubleshoot technical issues.
Critically validate AI-generated analyses, calculations and recommendations against underlying datasets and statistical evidence.
Integrate AI-assisted workflows into daily analytical activities while maintaining accuracy, consistency and sound analytical judgment.
Cross-functional collaboration and documentation: Work closely with Data Science and Customer Success teams to communicate investigation results and coordinate issue resolution.
Translate complex statistical and modeling concepts into clear, actionable recommendations for technical and non-technical stakeholders.
Track investigation queues, technical issues and resolution progress using Jira.
Maintain concise investigation reports, configuration change logs and technical documentation to facilitate collaboration, knowledge sharing and effective handovers.
Performance and continuous improvement: Work toward defined operational objectives, including reviewing new model alerts and documenting initial hypotheses within one business day.
Aim to identify root causes and provide recommendations within three business days of an investigation.
Maintain a high level of investigation accuracy, targeting at least 90% correct classification of data-related versus model-related issues.
Ensure every approved or rejected configuration change is supported by documented comparisons of model fit and business plausibility.
Progress toward independently managing the complete model alert investigation process within the first 60 days.