Time Intelligence
Suggest where the time went. Let people correct the story.
Help people understand where their day actually went.
Time Intelligence explored how location, calendar events and connected applications could become a timeline people could inspect and correct.
Different signals each describe part of the day.
Location, calendar entries and connected apps provide different views of work, travel and meetings. The product model brought them together and allowed overlapping activities.
An inferred day can be plausible and still be wrong.
Location tracking had a battery cost, and activity inference was still learning. A useful record therefore needed visible context and a way to correct its boundaries.
Build the correction model early.
As the sole designer, I worked with product, executives and engineers through weekly iteration. We used a Material UI kit and developed the interactions directly with engineers so the timeline and its corrections were tangible.
Connect signals to an editable interpretation.
The reconstructed model links connected sources, activity segments, inspection and category insights. A Replicon connection was part of the design for publishing the record to timesheets.
- Location
- Google Calendar
- Connected applications
- 01Assemble activity segments
- 02Inspect source context
- 03Split or merge boundaries
- An editable day
- Category insights
- Replicon timesheet connection
Inference → human correction → a more useful record.
Choose the signals that inform the day.
The design connects location, Google Calendar, and other supported applications. A Connections screen shows what is connected and offers settings and disconnection.
Make time visible as segments.
A timeline places activities in the day, including work, travel, and meetings. It must accommodate overlapping time rather than forcing everything into one sequence.
Show the context behind a segment.
Segment details include category, duration, start and end times, and location. The record is available for review.
Turn the record into a useful view.
Category insights summarize how time is distributed. A Replicon connection supports publishing the record to timesheets.
A reconstruction of the documented product model. It does not collect location data or infer your activity.
Make the source, segment and correction readable.
These wireframes explain the documented product model. The sample activities are illustrative; this page does not collect location data or infer anyone’s day.
Choose what informs the day
Make an inferred day readable
Put the boundary in the person’s hands
Reconstructed from the documented design; sample content is illustrative.
Let people disagree with the timeline.
Segment details expose category, duration, times and location. Split and merge controls let people correct an inferred block; the interaction below demonstrates that relationship with a sample day.
One inferred segment. The last hour was a meeting.
Illustrative split and merge interaction, redrawn from the source. The sample day is not a live activity record.
Project images coming soon.
Weekly team iteration made the open questions concrete.
The source describes R&D iteration, not a completed user study. Further evaluation should test correction effort, understanding of connected data and confidence in the record. Privacy and integration statements remain implementation requirements, not verified guarantees.
An R&D model for a correctable account of the day.
The presentation describes an application and a planned release. It does not establish a shipped date or measured productivity gain. The design contribution is making an inferred answer inspectable and editable.