Placemaker Tools

Overview

Think of it as a Wikipedia for places—an editor and moderation system for real locations: restaurants, stores, venues, and more.

Foursquare’s Places dataset powers maps, apps, ads, attribution, and recommendations. Automation alone cannot keep POI truthy. Placemakers (community contributors) work alongside Foursquare to add, correct, and validate place data—so the dataset reflects what is actually on the ground.

This project focused on making high-volume human review usable at scale: clearer queues, map-backed fixes, structured add-place, and visible contribution history—without turning moderation into a black box.

Placemaker Tools = the system Foursquare uses to crowdsource and validate real-world location data at scale.

I shaped product UX for contributor-facing flows—pairing with Engineering, Trust & Safety, and data operations so human review at scale stayed usable, fair, and aligned with dataset quality goals.

Work spanned end-to-end product narrative (onboarding → review → map fixes → contributions), interaction patterns for peer validation, and prototypes that became the shared reference for Engineering delivery.

Who it’s for

Three audiences share the same Places truth—but they show up with different jobs to be done.

Problem

The system continuously ingests tens of thousands of suggested edits. Without clear queues, map context, and community checks, noise wins: duplicates, stale hours, and wrong pins erode trust—and that trust debt propagates into every product that consumes Places.

Scaling review means more than adding headcount. AI agents and humans need shared surfaces—so triage is fast, decisions are explainable, and contributors still feel ownership.

The core design tension: optimize for throughput without sacrificing judgment quality or turning contributors into anonymous click-farms.

Contributor signals

Themes from contributor and ops conversations that shaped the product story.

Defining scope

Boundaries we used to steer design and engineering.

Approach

• Review queue first — Surface places that need verification (bad hours, missing fields, conflicting signals) so contributors know where attention matters.
• Map as truth-check — Let people see nearby context when fixing pins, duplicates, or closures—spatial mistakes are easier to catch on a map than in a table.
• Structured add-place — A guided form for new venues keeps categories, address, and hours consistent for downstream consumers.
• Contribution tracking — Show edits and impact so power users stay motivated and accountable.
• Community validation — Approvals, rejections, and voting turn individual judgment into shared consensus.
• Prototype in the product language — Cursor-built flows for onboarding, filters, and contributions validated interaction density before Engineering committed to delivery.

Product surfaces

Primary contributor journeys—jump between moments to see how onboarding, discovery, and accountability fit together.

Onboarding

First-run guidance that frames Placemaker work as quality stewardship—not freeform map editing—so new contributors know what “good” looks like before they hit the queue.

Key design decisions

Separate verification work from open-ended map browsing

The review queue optimizes for throughput and clarity; the map view optimizes for spatial fixes. Mixing both in one mode created cognitive overload—splitting them kept each task honest.

Schema-driven add-place, not freeform blobs

New locations funnel through structured fields (name, address, hours, categories) so automated checks, deduping, and downstream APIs get predictable shape—reducing “garbage in” at the source.

Transparent contribution history

Contributors see their edits and outcomes—building trust and making moderation auditable when disputes arise. History is a product surface, not a log buried in admin tools.

Peer review as a first-class state

Edits are not final until others weigh in where policy requires it—designing explicit approve / reject / needs info paths kept the social contract visible in the UI.

Room for AI-assisted triage without bypassing accountability

Surfaces assume humans and agents may share queues over time. Design prioritizes explainable states and human overrides so automation can speed triage without silently rewriting Places truth.

Outcome

Design direction converged on a single story: scale human judgment without drowning people in noise—with surfaces that internal ops, power contributors, and emerging AI-assisted workflows could share.

Impact