Testing our way to a 15% conversion lift at Grubhub

Built the Diner Retention design team from the ground up and led the core diner experience through Grubhub’s highest-growth period — A/B testing every feature we designed, for a 15%+ cumulative conversion lift with the App Store rating held above 4.8.

Role: Product Design Manager
Company: Grubhub
Year: 2017–2018

Context

Grubhub was in its highest-growth years, and the diner app — iOS, Android, and web — was where that growth was won or lost. I joined to build the Diner Retention design team, growing it from 2 to 6, and then took on the core diner experience with 10+ product and content designers.

Problem

Growth at that pace exposes every seam: three platforms drifting apart, features shipped on conviction rather than evidence, and a small team asked to cover the whole ordering journey from finding a restaurant to checkout and coming back again.

Approach

Test every feature we designed. Nothing shipped on opinion. Every design change ran as an A/B test against a measurable outcome — conversion, retention, rating — so the team learned what actually moved diners, and design earned a seat in roadmap decisions with numbers rather than mock-ups.

One design system across iOS, Android, and web. Established the shared system so the three platforms stopped diverging and the team could ship a change once and carry it everywhere.

Google design sprints, one or two a month. Made sprints a standing practice for the hardest problems, so discovery and validation happened with real diners before anything was built.

What we shipped

Grubhub diner homepage, signed in: a search bar with delivery or pickup, a saved work address and a cuisine field over a pizza hero; below it, Your orders with three recent restaurants and Preorder for later buttons. — Diner homepage, signed in. Search leads with delivery or pickup and a saved address, and “Your orders” brings recent restaurants back with one-tap preorder — retention built into the front door.

A Grubhub restaurant page for Sharetea with an Order Again section listing past drinks and their customizations, a note where an add-on is missing, and a cart panel with Invite friends and order together. — Restaurant page with “Order Again”: past items with the exact customizations you chose, a flag when a detail is missing, and a cart that invites friends to order together — reordering as the loyalty loop.

Four App Store screenshots of the Grubhub app: a restaurant list, restaurant offers with 25% off tags, a menu item with customization options, and a review-order screen showing subtotal, fees, tip and total. — App Store screens for the diner app: browse restaurants, restaurant offers, item customization, and a checkout that shows fees and tip before you place the order — the core ordering flow the team owned.

The winning tests clustered in four areas: profile, personalization, promos and loyalty, and reviews — each one a place where remembering the diner, or rewarding them, turned a visit into an order.

Outcome

15%+ cumulative conversion lift from winning A/B tests across profile, personalization, promos and loyalty, and reviews
App Store rating held above 4.8 through the growth period
A design team of 10+ product and content designers, a shared design system across three platforms, and a sprint cadence the team kept running

What I’d do differently

I was a fairly new manager then — three years in — and not as well-rounded as I could have been in how I coached designers and negotiated with my peers. The designers were so talented that it felt like we could do anything if we put in the hard work, and I led that way. In hindsight I would give more grace. When a hypothesis didn’t win, my instinct was to push straight on to the next solution; I would slow down instead and give a fast-moving team time to recharge.

What I carry forward

A/B testing every feature is what taught me to speak funnel with marketing and dollars with finance. It is the habit behind everything I’ve done since in acquisition design: earn the roadmap with evidence, and let the numbers, not the loudest voice, decide what ships.