Engagement Brief

Tony Tushar Jr.
Marketing Data Modeling.

Based in Minneapolis. I build the marketing data models inside your Snowflake environment that turn loyalty, sales, and behavioral sources into customer objects your marketing team can query and activate.

The problem this role solves

Loyalty, sales, digital behavior, app events, card data. Usually five different homes and no resolved customer across them. The work is the warehouse layer that resolves identity and shapes the result into customer objects marketing can query directly.

How to start on solving this problem

A typical first ask from marketing leadership: how is our churn changing, and can we flag customers likely to lapse so we can reactivate them?

Answering that end-to-end needs identity resolution, a rolling customer snapshot, an event history, a lapsed definition, and an audience the activation team can use. Two sample customers, walked through:

How I model

  • Identity spine first. Everything joins to a stable customer grain. Match rates get validated against real data before anything else gets built on top.
  • Fact, change log, event stream as the consumption layer. SCD2 customer fact for current and historical state. Change log for attribute-level deltas. Append-only event stream for behavioral signals.
  • Semantic layer as the shared definition. Dashboards, notebooks, and AI-assisted analytics all pull from the same certified measures and dimensions. One definition of a loyalty member, net sales, lapsed customer.
  • Audience tables as the activation handoff. Segments are queries against the model, materialized on cadence, with versioned membership.
  • Map to known standards where they help. Behavioral and identity entities map cleanly to XDM. Retail transaction entities map cleanly to ARTS ODM. Cuts schema invention.

Where I've applied this before

Performance marketing agency, 200+ brand clients. Identity resolution across eight sources moved audience match rates from 30 to 60 percent, reclaiming roughly $60K per month in paid search spend on a $200K monthly budget. Built the GA4 to BigQuery to dbt pipeline that served the full portfolio across paid, influencer, affiliate, and email.

B2B media company, lead data practitioner. Built the customer data foundation inside the warehouse for a CDP investment that had not been fully executed. Defined the event taxonomy, landed a semantic layer, and set up the modeling that carried segmentation, campaign attribution, and the AI agent work that followed.

How the first weeks would go

  • Weeks 1-2. Sit with marketing, loyalty, and analytics stakeholders. Capture the 10-15 business questions driving decisions. Profile the two priority sources, usually identity and transactional: grain, null rates, match rates, join readiness.
  • Weeks 3-4. Build and validate identity resolution. Land the first customer fact and change log. Demo against one of the harvested questions so stakeholders see a real answer coming out of the new model.
  • Weeks 5-6. Ship the first activation-ready audience, materialized and handed off. Bring in the second source using the same profile-and-map pattern.

More on the site. CV · Case Studies · Field Notes