To assist guests in finding items in stores being remodeled, Target pulled out all the stops.

One initiative involved standing up an SMS-based store search, which allowed users without the Target application to find items in the store.

Inventory searches would query the Target API for a product in a store and get back an aisle / adjacent location.

We then took that aisle & recursively searched the store's map (conveniently served as an SVG) to determine where the aisle was located in the store (e.g., back-left).

The service ran for 8 months in 47 stores and sent/received over 500,000 messages.

Under the hood, user messages were first parsed with Natural to determine intent, and then entities were extracted using spaCy.

Special Thanks

Thanks to Scott Vehrenkamp for pointers on natural language processing. The boy is one smart cookie.

In-store CTA

In-store CTA

Sample convo

Sample convo