Snapchat Context Cards

Surfacing context without requiring a search

The situation

In 2016, Snapchat acquired Vurb, the search startup where I was Head of Product. The thesis behind the acquisition: Snapchat had massive location data from snaps but no way to make it useful. Every day, millions of users were sharing snaps from restaurants, landmarks, and venues. That context disappeared the moment the snap was viewed.

The obvious solution would be search. Let users type what they're looking for. But that's not how people use Snapchat. They open the app to see what friends are doing, not to research where to eat. Adding a search box would fight the core behavior instead of working with it.

The question became: how do you surface useful context without asking users to do anything?

Finding the seam

I started by looking at where location data already existed in the product. Snaps had geotags. The Snap Map showed where friends were. Geofilters let users stamp their location on content. The infrastructure was there. What was missing was the bridge between "I see my friend is at a place" and "I want to know more about that place."

The insight was simple: the moment of curiosity happens when you're already looking at something. Not before. Not after. If a friend posts from a restaurant, that's when you want to know if it's good. If you see a snap from a landmark, that's when you want to know what it is.

We didn't need search. We needed context to appear exactly when it was relevant.

Building the integration layer

Context Cards pulled information from third-party partners and surfaced it inline when users swiped up on location-tagged content. No query required. See a snap from a restaurant, swipe up, get reviews and hours. See a snap from a city, get things to do nearby.

The partnerships were critical. We integrated Foursquare for venue data, TripAdvisor for reviews, Michelin for restaurant ratings. Each partner had different APIs, different data formats, different ideas about what mattered. My job was translating between what they could provide and what would actually be useful in a 3-second interaction.

The design constraint was brutal: whatever we showed had to make sense instantly. Users weren't going to read paragraphs. They weren't going to scroll through options. The card had to answer the implicit question before they consciously asked it.

Scaling through filters

Context Cards worked, but they were reactive. You could only get context on content someone else had posted. I wanted to flip it: let users add context to their own snaps.

Location Filters already existed, but they were generic. A "Seattle" filter didn't tell you much. We built venue-specific filters that pulled from the same partner data. Snap from a restaurant, get a filter with its name and rating. Snap from a museum, get a filter with what's currently on exhibit.

This created a loop. Users added context to their snaps, which made those snaps more interesting to viewers, which drove more swipe-ups on Context Cards, which drove more awareness of the filters. The product grew itself.

The results

Context Cards and Location Filters shipped to 200 million daily users. Filter engagement increased 75% after the venue-specific rollout. We launched partnerships with Foursquare, TripAdvisor, and Michelin, with more in the pipeline when I left.

More importantly, we proved that discovery doesn't require search. You can surface the right information at the right moment if you pay attention to where curiosity actually happens.

What I learned

Work with behavior, not against it. The search box would have been easier to build. It also would have been ignored. The best products find the seams in existing behavior instead of demanding new habits.

Constraints force clarity. A 3-second attention window sounds limiting. It's actually liberating. You can't overthink when you have no room for excess. Every pixel has to earn its place.

Partnerships are translation problems. Foursquare, TripAdvisor, and Michelin each had years of data and strong opinions about how it should be used. The product work wasn't just what to show — it was how to take what they had and reshape it for a context they'd never designed for.

See all work · Read about Teams · Read about Segment · Email me