Establishing Design as a Strategic Partner

The first challenge was getting design involved early enough to matter. The BE/ML teams weren’t accustomed to having a designer embedded in technical architecture decisions. I had to demonstrate that design input at the system level would save time and rework downstream.

I started by creating foundational diagrams—visual representations of how the system actually worked, translated from engineering complexity into stakeholder-legible concepts. These became the shared mental model that grounded every cross-functional conversation.

[Image: Foundational System Diagram]

The system diagram I created to align cross-functional teams on an understanding of P+ strategic place in the platform, higher-level UX implications, and where it may fit in the advertisers mental model.

Navigating the Automation vs. Control Tension

The core design challenge was philosophical before it was visual: how much should the machine decide versus the advertiser?

I developed the Optimal Setup Framework—a system that maps advertiser goals and spend levels to intelligent defaults. Instead of exposing dozens of configuration options, we ask advertisers what they’re trying to achieve. The system translates intent into configuration. This preserved advertiser agency without requiring expertise.

[Image: Optimal Setup Framework]

The Optimal Setup Framework: mapping advertiser intent to system configuration.

[Image: P+ Campaign Creation Flow]

The P+ campaign creation flow—simplified setup that preserves advertiser agency.

Designing for Trust During Uncertainty

Automation fails silently. When a P+ campaign underperforms in its first week, the advertiser doesn’t know if the system is broken or still learning.

I proposed the Learning Phase indicator—UI that sets expectations during the optimization period. When a campaign is new, we explicitly communicate: “Your campaign is in its learning phase. Performance may fluctuate as the system optimizes. We recommend waiting before making changes.”

This wasn’t just a label. It was a behavioral intervention. By naming the phase and setting expectations, we reduced premature pauses—protecting revenue and giving the system time to optimize.

[Image: Learning Phase UI]

The Learning Phase indicator—setting expectations to prevent premature campaign pauses.

Driving Cross-Functional Alignment

P+ touched Growth, Ads Quality, Monetization, Engineering, PMM, and Sales. Alignment required more than good designs—it required creating forums for coordinated decision-making.

I initiated the weekly Bidding Meeting—a cross-team sync coordinating P+, Bid Multiplier, ROAS Bidding, and LTV signals work. I was also a key contributor to the “Triple-A” adoption strategy—the go-to-market plan for rolling P+ to advertisers.

Key Decisions I Drove

Optimal Setup over manual configuration. Instead of exposing every lever, we ask advertisers what they want to achieve and configure the system for them. This reduced setup friction, improved adoption, and preserved the feeling of agency without requiring expertise.

Learning Phase indicator shipped to 100%. Advertisers were pausing campaigns prematurely because they didn’t understand early performance fluctuations. The indicator set expectations and protected revenue.

Fine-tuning as a separate layer. P+ is full automation, but some advertisers want more control. Rather than compromise P+ with manual overrides, I advocated for Value Expression and Bid Multiplier as the “fine-tuning” layer—separate capabilities that work with P+, not against it.

Foundational alignment before feature design. I invested early weeks in diagrams and frameworks rather than UI. This felt slow but paid off: the shared mental model reduced thrash and accelerated every subsequent decision.

Impact

~10% CPA improvement

$1M/week incremental revenue

$523M platform revenue (~33% of daily lower-funnel spend)

$104M Q4 2024 contribution (combined with Bid Multiplier)

100% GA rollout

CAPACITY BUILDING

Shipping P+ revealed gaps in the platform’s ability to guide advertisers when they deviated from optimal setup. I scoped and initiated adjacent workstreams.

Mid-flight recommendations. If an advertiser opts out of P+ defaults and makes suboptimal choices, the platform should intervene. I developed the framework for performance recommendations that quantify payoff tied to stated goals—showing advertisers what they’re leaving on the table.

Campaign status as entry point. I broadened the status indicator in the reporting table from a health signal to an information center surfacing when campaigns need attention—creating a pathway to recommendation adoption.

Guidance Framework. I led the effort to classify guidance types (descriptive vs. prescriptive vs. guardrails) and define the advertiser journey from awareness through adoption. This positioned the Guidance team to scale without reinventing the wheel.

Where This Fits

Performance+ is the “full automation” layer of an Advertiser Automation Spectrum I’ve been building at Pinterest:

• Full Auto (P+): Set a goal, we handle the rest

• Fine-Tuned (Value Expression, Bid Multiplier): Express what matters more

• Recommendations (Learning Phase, Guidance): Here’s what we suggest

• Manual: You control everything

Related work: Value Expression · Bid Multiplier

← Back to Work

Navigating the Automation vs. Control Tension

The core design challenge was philosophical before it was visual: how much should the machine decide versus the advertiser?

I developed the Optimal Setup Framework—a system that maps advertiser goals and spend levels to intelligent defaults. Instead of exposing dozens of configuration options, we ask advertisers what they’re trying to achieve. The system translates intent into configuration. This preserved advertiser agency without requiring expertise.

[Image: Optimal Setup Framework]

The Optimal Setup Framework: mapping advertiser intent to system configuration.

[Image: P+ Campaign Creation Flow]

The P+ campaign creation flow—simplified setup that preserves advertiser agency.