Unearth Campaigns · Executive dashboard · Privacy compliance
I led an executive dashboard that turned 1.7 million state-mandated deletion requests into a clear answer: which audiences we lost, and what to do about it.
All data shown is placeholderA new state privacy law began removing people from our audiences, and leadership couldn't see the impact.
Learned what leadership needed, defined the metrics and layout with the developer, set impact criteria with the PM and pitched an AI summary.
A high-level overview leadership can scan in two minutes, with a daily AI summary and recommended next steps.
Methods
The problem
California's Delete Request and Opt-out Platform (DROP) lets residents send one deletion request to every registered data broker at once. As a registered broker, our company has to pull new requests at least every 45 days, delete every match and report the status of each request back to the state. Missing the deadline costs $200 per request, per day. Every deletion also shrinks the audiences our product sells to clients.
Leadership kept asking the same questions, and nobody could answer them without pulling the data by hand.
What I did
I mapped each question to a part of the report: who was removed and how, which audiences lost the most and the raw data for anyone who needed to dig in. A contract developer built it; I owned what it showed and how it read.
| Audience | Removed | % lost |
|---|---|---|
| Audience A | X,XXX | XX% |
| Audience B | XXX | XX% |
| Audience C | X,XXX | X% |
| Audience D | XX | X% |
I iterated on a beta with the developer first. Before the first cycle closed, I walked leadership through a demo, gathered their questions and met with them to understand their concerns about DROP and what they wanted to see. Their answers became priorities for the whole report, not just fixes. The biggest lesson: percentages matter more than raw counts.
One-click sign-in with work accounts and sortable tables, so anyone on the team can open it and find what they need.
Company branding and a mobile pass, so the report holds up on a phone.
Faster loading, date filters and exports, so leadership can check one delivery or the whole cycle.
Leadership cared most about how much of each audience was lost, so tables sort by % lost and the funnel shows rates.
An AI summary at the top calls out what matters most in the data and recommends what to do next, refreshed daily.
Leadership needed the story, not the spreadsheet. I proposed a summary written by Claude from the report's own figures, then edited how it reads: full width, no repeated descriptions and our team's words ("audiences," not "lists").
AIWritten by Claude from the figures in the tables below. Check a figure against its table before acting on it.
The disclosure line was a deliberate trust decision: leadership sees that the summary is AI-written and where to verify it.
Before the report could recommend next steps, we needed shared rules. I brought three questions to the product manager: What % of an audience lost counts as low, moderate or significant? Should that change with audience size? And what should each level trigger?
| Impact level | % of audience lost | Next step |
|---|---|---|
| Low | Under X% | Monitor next cycle |
| Moderate | X% to Y% | Refresh the audience |
| Significant | Over Y% | Rebuild, and flag affected campaigns |
Once leadership was using it, I mocked up a new top section that answers "how bad is it and are we on time" before anything else.
| Audience | Removed | Impact |
|---|---|---|
| Audience C | X,XXX | Significant |
| Audience F | X,XXX | Moderate |
| Audience E | XXX | Minimal |
| Audience | % lost | Removed |
|---|---|---|
| Audience A | XX% | X,XXX |
| Audience B | XX% | XXX |
| Audience C | X% | X,XXX |
Drag to compare, or choose a change below to see it.
Key decisions
Results
deletion requests from the first state cycle, visible in one report
release rounds in 4 weeks, each driven by leadership feedback
AI summary refresh, so leadership never reads a stale number
What's next
Leadership lives in Slack. Slack's new Surfaces let Slackbot build a live dashboard from connected data and pin it to a channel, where people can filter it, comment and ask follow-up questions. I plan to explore delivering the report there instead of asking leadership to open another tool.
Here's this week's DROP impact. 2 audiences crossed the refresh threshold.
Which client campaigns use those 2 audiences?
Lock the thresholds with the PM, then bring the labels and recommended next steps back into the summary.
Post to Slack when unreported requests get close to the state deadline, or when an audience crosses a threshold.
DROP keeps running, so show whether each audience is stabilizing or still shrinking.
Once leadership decides the policy, show which client campaigns are affected so account teams can reach out first.
From spreadsheets to grant-ready data for a shelter and service provider, with reporting built into the work staff already do.
Read next