Visit Essentials

Visit Essentials enables you to know what kind of locations your users are visiting using Venue Type Mapping and Home/Work detection models. Additionally, using Semantic Time, it tells when the place was visited relative to the user's personal timeline.

• Home/Work detection - Home and work locations are identified based on visit patterns and usually become available during the first week of using the SDK.

• Venue Type Mapping - Based on the location, time of the day, and other features, it estimates if the user is going for drinks, sports, education, shops, etc.

• Semantic Time - The definitions of morning, lunch, evening and night are personal and necessary for profiling users and knowing when it’s the right time to engage them.

Combining these three components enables identifying morning commutes, lunch breaks, longer-than-expected stays at work, etc.

Note that we do not expose exact location and instead use Venue Type Mapping (knowing only the type) mainly for privacy reasons. Knowing the type is sufficient to conclude the lifestyle while preserving privacy. Other benefits include increased accuracy since we can tell with higher certainty that a user is in a bar instead of pinpointing the exact one (if there are a couple nearby).

Visit Essentials works on-device and has low latency, thus enabling a more fluent “when to engage?” experience. For example: sending the message the moment the user exits a work location at lunchtime. To construct venue candidates, we download map tiles to the device, but these span 39x39km, thus the backend never receives the precise location that a user is visiting.

Predictions from Visit Essentials are returned each time the user enters or exits some location.

Frequently Asked Questions (FAQs)

I moved to a new home (or: I changed jobs recently), why are my detections wrong?

Users can have multiple work and home locations, but a couple of repeated visits are necessary until the model can make certain predictions. After changing your home/work location usually one week is sufficient to see correct detections.

During vacations will a hotel be identified as home? What happens with the current home location?

Since we support multiple homes it is expected that the hotel location can potentially be identified as a second home after a couple of overnight stays. This does not imply that the model forgot about your actual home, when you come back from vacation your old home will be identified correctly on the first visit.

Why is my work location not identified correctly?

Note, that if you have an irregular office schedule it might confuse the model. In such a case, a couple of weeks might be needed to gather enough data to make a reliable prediction.