Nextdoor · Design Systems · 2023

Business Experience

Lifting business recommendations 15% with AI-powered discovery

Company: Nextdoor
Role: Staff Product Designer
Duration: 2023
Scope: Business discovery, recommendations, and bookmarks across Home Feed, Map, and Pages
Team: Engineering Manager, Product Manager, Product Designers, User Researcher

A business recommendation card in the Home Feed — Home Feed.

A redesigned Nextdoor business page — Business page.

A neighbor recommendation thread for a local business — Recommendation thread.

The recommendations summary for a business — Recommendations summary.

Reframing the opportunity

Word-of-mouth needed to be surfaced in posts and comments, so the community could build relationships with businesses and each other.

76% of neighbors said community shaped their purchases. When the legacy tagging flow that surfaced those recommendations broke, engagement dropped 42% — proof of how much weight that word-of-mouth was already carrying, on a platform reaching 1 in 3 US households.

How might we help customers show support for local businesses so they thrive?
How might we help businesses prove their reputation so they grow new customers?

A legacy Nextdoor business page for a handyman service — Legacy business page.

A legacy business page scrolled to its Similar Businesses section — Similar businesses, legacy layout.

A legacy Nextdoor business page for a babysitting and tutoring service — Legacy business page, another category.

Key design decisions

01Surface word-of-mouth contentMake user feedback visible and discoverable to build customer trust with local businesses.

02Optimize for actionsMake it easy for customers to find and share the businesses they love.

03Reduce participation frictionLet users discover, save, and share businesses with ease through familiar patterns.

Product & design strategy

Four guiding pillars that became the north star for every tradeoff and stakeholder conversation.

01Fix broken taxonomiesRepair recommendation pathways that became broken after system changes.

02Improve core usabilityMake it intuitive for users to find, evaluate, and take action on businesses.

03Provide path to sharesCreate more natural opportunities to save and share businesses using universal interaction patterns.

04Improve performanceAddress the latency and reliability issues that blocked engagement.

A jobs-to-be-done map for neighbor and business needs — Jobs to be done — mapping neighbor and business needs across Home Feed, Map, and Pages.

Redesigned business page

AI-generated recommendation summaries, pulled from verified neighbor reviews

Our key metric was to increase engagement, growth, and raise awareness around businesses. Research showed that high engagement occurs in comments when users request services. I partnered with the machine learning team to identify posts seeking services, which then enabled the system to detect and auto-tag mentioned businesses on classified posts.

The redesigned business page, with AI-generated recommendation summaries and a sticky nav

Auto-tagging detecting a business mentioned in a classified post

Lead vision, systems, and team

01Vision & StrategyDefined the product vision and ran principles workshops with designers and PMs, giving shared language for every tradeoff.

02Design SystemA/B tested the business card for visibility and feed depth, consolidating three variants into one component reused across feed, comments, and pages.

03TeamMentored two designers, settled stakeholder debates with shared principles, partnered with trust and safety on abuse prevention.

Sketches exploring the design principles workshop — Principles exploration.

The finished design principles poster — The design principles, settled.

Role & scope

Strategy

Defined product strategy and design principles
Illustrated Jobs to be Done to align teams
Partnered with PM on success metrics

Execution

Led end-to-end design across phases
Owned all shipped screens for business pages
A/B tested business card components
Partnered with ML team on business & recommendation detection

Leadership

Mentored two product designers
Led design principles workshop
Resolved stakeholder debates with shared principles
Coordinated with trust and safety on abuse prevention

The business card component, consolidated to one and reused across feed, comments, and pages — The business card — A/B tested, then consolidated from three variants into one component reused across feed, comments, and pages.

Results

Outcome	After
Total recommendations	+15%
In-comment recommendations	+84%
Comment bookmarks	+22%
Home Feed bookmarks	+9%
Page bookmarks	+14%
Share rate	+2%

The +84% reflects in-comment engagement including business owner self-mentions, an edge case the ML classifier had not modeled — total recommendations lift was +15%. I scoped a fix with the vitality team to filter self-mentions before transitioning off.

← All projects