Palate
Mobile app with a taste model learned from behaviour, location aware search, a social layer, and recommendations that work from the first open.
Palate was a founding team with a strong product thesis about how people actually choose where to eat, and nothing built yet. The product had to be good enough that users would open it repeatedly, in a category where the incumbents are very large.
We built the mobile application and the services behind it.
A taste model that learns from what users actually do, not from a questionnaire they will abandon. Rating a few things reveals less than opening and scrolling through dozens.
A recommendation engine blending collaborative signals with content attributes, tuned so a user with no history still gets useful results. The cold start problem is what kills discovery apps. Palate's version had to work from the first open.
Location aware search with geospatial indexing, fast enough to feel instant on a mobile connection.
A social layer: follow friends and taste matched strangers whose feed consistently surfaces things you end up trying.
Image handling with compression on the device and progressive loading, because the app is full of photos and users are on mobile data.
Offline caching so the app opens to something useful on a bad connection.
Instrumentation throughout, so the product team can see which recommendation strategies actually drive return visits, and tune without an engineer.
Two to three engineers from MVP through launch and into ongoing support.
A shipped Palate app with a recommendation system the team can tune from real usage without an engineer.
Capabilities used
Tell us about it. We'll tell you honestly whether we're the right team to help ship it.