Data Engineer, Marketing Science

Power Digital is an AI-Native Growth Firm built at the intersection of world-class talent and proprietary technology. We help leading and emerging brands grow by combining deep vertical expertise, strategic thinking, and enterprise intelligence to drive measurable business outcomes.

For more than a decade, we have built our business around one belief: the best technology makes great people better. Our teams of strategists, creatives, analysts, and technologists use Omega, our AI operating system, to surface opportunities faster, investigate what is driving performance, and execute approved work with greater speed and precision. Behind Omega is Iris, our proprietary intelligence layer, built on years of real performance data, strategic frameworks, testing, and learnings from across our client portfolio. AI expands the capacity of our people to think, create, solve harder problems, and make better decisions. Human judgment, creativity, and accountability remain central to how we work. 

As a people-first firm, we value diversity in backgrounds and experiences. We strongly believe our people and culture are key to our success. We partner closely with our clients as an extension of their teams, working across strategy, execution, and measurement to identify what is truly blocking growth and determine what to do next. Our culture is built on people who are curious, ambitious, accountable, and committed to delivering meaningful results for our clients and for one another.

At Power Digital, we believe the future belongs to teams that combine the best talent with the best technology. We are building that future today.

You'll sit on the fusepoint Marketing Science Team and own the data behind Power Digital's measurement products: Media Mix Modeling (MMM) and incrementality testing (Matched Market Testing, MMT). Our measurement tools are used by internal and external teams to read model results, plan budgets, pick test markets and report on incrementality tests. Your job is to make sure the data needed for measurement analysis is clean, reconciled and ready for modeling, so the analysts and data scientists spend their time on modeling, not cleanup.

You'll work closely with the Data Team to support, maintain, and engineer data sources related to measurement and marketing metrics. 

• →Own the MMM data layer in Snowflake. Build and maintain model-ready daily and weekly marts for each client that combine paid media (spend, impressions, clicks by channel and campaign), ecommerce revenue and orders (new vs. returning, gross and net, refunds, discounts), site and email/SMS data from GA4, and any non-media drivers. Every mart has consistent channel groupings, currencies and timezones.
• →Own the MMT geo data layer. Build geo-level media and sales tables at DMA (or Comscore), state and zip level for every major platform (Meta, Google, TikTok, Pinterest, Amazon, Microsoft, Programmatic, etc.) and for CMS/CRM/Amazon/Retail Datsets, with standard market names that line up across sources. These tables feed automated market selection, pre-period validation, synthetic control models and test readouts.
• →Build and automate a data quality process. Turn the checks we currently run manually into automated Snowflake/dbt tests and alerts: date gaps, zero-spend or zero-revenue days, cost without impressions (and the reverse), duplicated or unmatched markets, outlier spikes, restatement drift and source freshness. Problems should be flagged before a model is trained or a test is read out.
• →Own and maintain model logs and exported model outputs (contributions, response curves, adstock, optimization results) well structured, versioned and easy to query.
• →Support onboarding of new measurement clients. Own and maintain a repeatable, mostly automated process: map their data feeds, check history depth and quality, and flag gaps before modeling starts.
• →Handle client-specific data needs as they come up. You are not responsible for client workflow, but you may need to jump in to assist the team as things come up.. Write client-specific overrides, channel mappings and marts when a client needs them, and decide whether each piece of logic belongs in the shared core or in the client's own layer.
• →Use AI coding tools to work faster (Claude Code, Cursor, Copilot, Hex's AI agent), and document the patterns that work so the team adopts them.

• 3+ years in data or analytics engineering, including 1+ year owning a production dbt project of meaningful size.
• Advanced SQL and Python, with production-grade pipeline code.
• Deep dbt expertise: incremental models vs. full refreshes, Jinja and macros, tests, snapshots, source freshness, exposures, and keeping a large DAG under control.
• Strong Snowflake skills: semi-structured/VARIANT data, dynamic per-client queries, query and cost tuning, permissions across separate client databases.
• Hands-on marketing and ecommerce data experience: ad-platform APIs, UTMs, attribution windows, platform-reported vs. warehouse-reported conversions, Shopify order and refund logic.
• An understanding of what measurement models need from data: a continuous time series, consistent granularity, stable channel definitions, enough history, and the reasons one bad week or market can skew results.
• Experience building multi-client data models, with the judgment to know when logic belongs in the shared core and when it belongs in a client layer.
• Git and CI/CD with automated testing you trust.
• Real adoption of AI-assisted development workflows.
• Comfort shipping iteratively based on feedback from analysts and app users.

Helpful but not required

• Direct MMM or incrementality experience (Meridian, Robyn, PyMC-Marketing, synthetic control/causal impact, geo-lift).
• Experience with geographic data: DMA/zip mappings, population and market-size data.
• Hex experience (SQL/Python notebooks, semantic models, published apps), or similar BI/notebook tools.
• Experience at an agency, consultancy or martech company.

• MMT coverage: Geo-level MMT tables for all data sources are created and in use by the Measurement team in the first 3 months of the role
• Measurement-ready data: Within 6 months, ≥95% of active MMM and MMT clients have automated, tested Snowflake marts feeding Hex, with no client-specific cleaning logic left in app notebooks for core sources.
• Data quality: ≥99% completeness and accuracy on critical measurement fields. Zero model refreshes or test readouts delivered with an undetected date gap, duplicated market or unreconciled revenue source.
• Reliability: ≥95% success rate for scheduled measurement pipelines, excluding documented vendor outages.
• Speed and efficiency: Cut new-client measurement onboarding time (data mapping and validation) by 30% within 12 months. Measurably reduce load and refresh times for the main Hex measurement apps.
• Client requests: Deliver ≥90% of client-specific measurement data requests within the agreed turnaround
• AI-assisted development: At least 1 reusable AI-assisted development pattern documented within 120 days and 2 within 12 months, each used in at least 2 production workflows.

• Deliver clean, model-ready MMM and MMT data in Snowflake so the measurement team’s data can be trusted without extra cleanup.
• Build a process to catch data issues before they reach a model or test readout.
• Build a single, documented measurement data foundation that grows with the client list and is ready for AI use.