Verve — Ad Marketplace Analytics

Product Analyst Intern — Verve, programmatic advertising marketplace

Role

Product Analyst Intern

Timeline

May 2025 – Jul 2025

Context

Verve, programmatic advertising marketplace

Type

Product Analytics / Data

Problem

Verve's leadership was making marketplace decisions without reliable, consolidated visibility into what was happening across millions of daily ad auctions. KPIs like CTR, eCPM, win rate, and margin were tracked in silos — no single view of supply-demand health, no automated monitoring, and reporting workflows that required significant manual effort every cycle.

Context

Verve operates a programmatic advertising marketplace where supply and demand dynamics across auctions directly determine revenue outcomes. The data infrastructure existed in BigQuery, but consolidating it into actionable visibility required both engineering work and deliberate product thinking about what to surface and how.

Why It Mattered

In ad marketplaces, auction dynamics change fast. A shift in win rates or eCPM that goes unnoticed for a week can represent significant revenue impact at scale. The team needed daily visibility — not weekly reports and not manual SQL queries — to make fast decisions about supply sourcing, pricing, and product priorities.

My Role

I owned the dashboard end-to-end: metric selection, SQL + Python pipeline development in BigQuery, Tableau build, and ongoing analysis. I also collaborated directly with PMs and engineers to translate the patterns I found in auction data into product requirements.

Constraints

Internship timeline. Data spread across multiple BigQuery tables with complex join logic. Needed to serve both executive-level summary views and analyst-level drill-downs in one interface.

What I Did

Built and launched a Tableau Health Dashboard consolidating 10+ marketplace KPIs — CTR, eCPM, win rate, margin, and others — giving leadership daily visibility into supply-demand health across millions of ad auctions.

Developed automated SQL + Python pipelines in BigQuery to monitor revenue trends and auction outcomes continuously. This replaced manual reporting workflows and made the data available at a cadence that actually supported fast decisions.

Analyzed auction dynamics across thousands of daily transactions to surface demand and supply-side patterns. When I found something meaningful in the data, I didn't just flag it — I worked with PMs and engineers to translate it into concrete product requirements and roadmap input.

Structured the dashboard deliberately: an executive summary layer with the critical numbers, and drill-down views for the analysts who needed to investigate further. Both groups using the same tool, with entry points designed for how each actually works.

Key Decisions & Tradeoffs

Decision 1

Chose Tableau over a custom-built dashboard given the timeline and the team's existing familiarity with it — the right tool for the actual situation, not the most technically interesting one.

Decision 2

Automated the pipelines first before building the UI. An automated feed that's slow to update is still more useful than a beautiful dashboard built on stale data. Getting the data infrastructure right was the foundation.

Decision 3

Worked directly with PMs and engineers when analysis surfaced patterns worth acting on — not just producing reports, but translating findings into product language. The goal was decisions, not metrics.

Outcome

Tableau Health Dashboard launched and adopted by leadership for daily marketplace monitoring. Automated BigQuery pipelines eliminated manual reporting overhead. Auction analysis contributed to product requirements and roadmap prioritization discussions with PMs and engineers.

Reflection

“Ad marketplace analytics taught me that the hard part of data work isn't the SQL — it's knowing what to measure and why. Picking the right 10 KPIs out of 50 possible ones, designing the dashboard so it answers questions rather than just displaying numbers, knowing when a pattern is worth escalating versus just monitoring — that's the judgment that actually matters.”

←Back to all work