Advertisers were asking for controls that would have broken the automation that was working.
Role
Senior Product Designer to Staff Product DesignerDesign lead, Value Expression · set the direction, coached the designer who shipped it
Company
PinterestMonetization Org
Timeline
Q4 2022 to Q2 2025Active development Q3 2024 to Q2 2025
Impact
$104M Q4 2024Combined with Bid Multiplier
01The stakes
Q4 2022
Performance+ was live and beating targets when the requests started.
Hard constraints fragment the automation's ability to optimize: every audience excluded shrinks the pool it can target, every spending cap limits its runway.
Performance+ was working. Then advertisers started asking for things that would break it.
Cap retargeting spend at 30%. Exclude California. Limit delivery to five audience segments. Each one had a real business behind it. Margins differ by customer type, inventory is finite, some regions carry more weight than others. Each one would also have made their own campaigns perform worse.
Advertisers don't actually want constraints. They want to express what matters more.
Where all three products came from
02The strategy
2023 to 2024
The “signals, not constraints” language became how PM, Eng, and PMM talked about the product.
Advertiser control versus automation performance was a standing argument. Value Expression is an implementation of the Optimal Setup Framework I wrote, the same framework Performance+ was built on. The framing settled it, and a mid-level designer took the work from there.
Framing
Established the framing the org adopted; it settled the standing tradeoff between advertiser control and automation performance.
Coaching
Ran design reviews as a mid-level designer moved to owning the work. That designer shipped Bid Multiplier, then took on Audience Weight and Spend Preference independently.
Principles
Lean into value: importance in relative terms, not absolute budget figures. Simplicity over precision. Expectation setting: a signal weights delivery, and the system still decides where the money lands. Reporting looks backward at what happened.
Prioritization
Studied how Meta handles Advantage+ audience definition and budget allocation, then led prioritization with PM: Bid Multiplier sequenced first, Audience Weight and Spend Preference mapped as follow-on releases.
Fig. 01 · Before. The Performance+ targeting controls, limited to country level, with region or state exclusions. Intent arrives as hard rules that cannot be softened.Fig. 02 · After. Performance+ targeting with value expression that guides how the system bids, with support for exclusions.
Fig. 03 · The reframe. Constraints become signals: the foundation for Bid Multiplier, Audience Weight, and Spend Preference.
Traditional model
Constraints
דOnly target California”
דCap retargeting at 30%”
דExclude these audiences”
Fragments delivery
Shrinks what the model can learn from
→
Value expression model
Signals
✓“Prioritize California”
✓“Weight prospecting higher”
✓“These audiences matter more”
Guides delivery
Gives the model more to learn from
03Shipped
Q3 2024 to Q2 2025
Advertisers read bid adjustments as spend guarantees; auction dynamics make spend unpredictable.
Advanced controls stay available but not foregrounded.
What an advertiser cares about stays the same from campaign to campaign. So the account holds the value rules, and campaigns inherit them without anybody restating anything.
The hard part came after. An advertiser sets a signal and hears a promise about spend, and the auction cannot make that promise. Warning states catch the extreme settings. Reporting shows what the preference actually did, next to live campaign performance.
The alpha shipped with blank percentage fields. Advertisers guessed at what to type. The team called it a “button of hope.” The brief for the next product had already named the risk: high potential for confusion, and no benchmark for how much deviation advertisers would tolerate. I wrote that down and shipped anyway.
Fig. 04 · Settings level. Account-level value rules for locations, products, ad placements, and devices. They define an advertiser's way of doing business and inherit across campaigns.Fig. 05 · Reporting reflection. The value rules surface back in Ads reporting, set against live campaign performance, so an advertiser can see the preference at work.Account and campaign figures shown are demo data.
Fig. 06 · Expectation setting. The hard part, mapped: a signal reads as a preference, not a cap; reporting shows what happened, not a forecast; warning states catch extreme settings.
$104M
Q4 2024, with Bid Multiplier
3
Products from one framework
Org-wide
Framework adopted
0
ML optimization fragmented
04What it enabled
2025 → 2026
The same signal model kept extending: audiences, product groups, budget allocation, reporting.
Audience Weight · the same signal model applied to customer lists and custom audiences.Shipped
Priority Products · product-group-level bid prioritization for margin, inventory, and seasonality.Backend alpha
Spend Preference · prospecting versus retargeting allocation as a signal rather than a hard budget split.2026 roadmap
Reporting clarity · a bid weight indicator shows how preferences translated to delivery outcomes.2026 roadmap
The part I'm proudest of is the handoff. A mid-level designer walked into a new product category with almost nothing settled in it, and needed direction to start. Bid Multiplier shipped with me in every design review. Audience Weight and Spend Preference, they took on without me.
