Case Study 01

Marketer Workbench

Unifying scattered ML tooling into a single workflow for growth marketers.

Role: Sr. Product Manager
Year: 2025-2
Team: PM, 6 eng, 1 design
Focus: MLOps · Workflow

Problem Definition

Marketers want to prioritize their best customers and prospects for offers and loyalty-building communications, but they can’t build the AI audience models needed to do this at scale.

Research & Discovery

Interviewed Customer Advisory Board and reviewed usage of existing Zeta built audience models. Lastly aligned internal AI Agent MCP/Orchestration Layer development.

Strategic Solution

Constrain to the most highly utilized types of models (LTV, Churn, and Click-through rate as an intermediate step towards Conversion). Stabilize metric definitions so that input variable definition is selection from known options, not writing from scratch. Leverage existing Zeta Social Graph data and data engineering platform.

Execution & Prioritization

Used MoSCoW to identify priority stages of simplifying our existing ML model training pipeline to leverage AutoML, extending data integrations to prepare for more custom client data, and only then integrated a chatbot-led interface that sits on top of our company wide Athena agent.

Impact & Metrics

Since Marketer Workbench’s ICP is an enterprise client on an all-encompassing contract, impact was measured by adoption and eventually by contract renewal influence (not just churn, but also size and terms). At launch, we had 4 commitments from early adopters, and within six months of general availability, Marketer Workbench became part of most enterprise contract negotiations.