Audience Management · Hew Suber

01The stakes

2019

Three versions of Audience Manager lived across four URLs.

Enterprise advertisers ran tens of thousands of audiences powering millions in revenue. The product could not keep up.

Custom Audiences and Lookalikes were where enterprise money got aimed. An advertiser holding tens of thousands of them had no way to search them, group them, or see which ones were in use. So they lost hours rebuilding audiences they already had. Or they made a second copy of one that existed, because nothing on the screen told them where the first one was running.

Before · three surfaces, no visibility

Three versions of Audience Manager
Spread across four URLs
No view of where an audience ran
Filters present but unused
Naming conventions doing the tool’s job
Defensive copies piling up

After · one surface, usage visible

01One unified Audience Manager

02Every audience mapped to campaigns and ad sets (groups of ads sharing a budget)

03Search by name, label, and event

04Cross-team and partner usage in view

Reuse became safer than copying.

When advertisers build their own system around yours, yours isn't working.

The failure signal that set the priorities

02Two tracks

2019 to 2020

Quant showed people bouncing between the Audience page and the Ad Set page. Filters tested well when prompted. On their own, people left them alone.

Nobody knew the filters were there.

Two tracks, one surface. Show where an audience was running. Make it findable.

I pulled the research together, decided what got built first, and designed both tracks at once.

Usage panel

A panel that put any audience next to the campaigns and ad sets running it, with the dates, the delivery status, and anyone else on the team or at a partner agency using it.

Search

Cut the filter types the product did not support. Added Events filtering, with the standard events plus a catchall for custom ones. Moved search up where people would see it. Rewrote every filter label so it says what the filter is for.

Consolidation

Three versions of Audience Manager, spread across four URLs, became one. The new surface ran on FDS, Facebook's shared design system, and its components moved toward Geodesic, the pattern set coming next.

The vision

I wrote where audience management should go next: performance insights tied to campaigns, search by behavior, bulk actions, lifecycle indicators. The document outlived the project and stayed the roadmap reference.

Audience Usage Panel mapping an audience to its campaigns. — **Fig. 01 · Audience Usage Panel.** Any audience mapped to the campaigns and ad sets running it, with delivery status and dates, without leaving the page.Account and campaign names shown are demo data.

Usage detail view with campaign-level breakdown. — **Fig. 02 · Usage detail.** Campaign-level view with ad set counts, delivery status, and scheduled windows, plus a History tab for the audience itself.Account and campaign names shown are demo data.

03Shipped

2020

Two directional success signals: less context-switching between the Audience and Ad Set pages, and fewer total audiences per account over time.

Advertisers could finally answer basic questions about their own audiences.

Is this audience in a campaign right now? How many ad sets does it power? Did a colleague or agency partner already use it? If I change it, what breaks? The Usage Panel answered all four on the page, with the first view of who else across teams and partner agencies had that audience running.

Search exploration with typeahead and time-based queries. — **Fig. 03 · Search exploration A.** Typeahead matching audience names and labels, with time-based queries detected in place and a date range picker opened when one is entered.Account and campaign names shown are demo data.

Search exploration with a progressive search-or-filters bar. — **Fig. 04 · Search exploration B.** One search-or-add-filters bar with progressive states: closed, first use, open filter, and applied selection.Account and campaign names shown are demo data.

3 → 1

Platforms consolidated

↓

Audience duplication

↑

Search and filter adoption

04What it changed

2020

The wins stayed directional by design: define the signals, watch the direction, skip the headline claim.

The most expensive behavior in the system was defensive duplication. Advertisers weren't making bad decisions. They were making rational ones, given what the product showed them, which was almost nothing.

We judged the work on what advertisers did next. Fewer trips over to the Ad Set page to answer a question. Fewer copies of the same audience piling up in one account. Neither of those makes a good slide.

Three platforms became one.