Graph Analytics
This is a work-in-progess project for building privacy-first, scientifically rigorous front end analytics based on graph data structures.
The basic outline of the project is a collector script that sends event data to an endppint, which stores the event as a set of triple statements in a triplestore. A front-end application can then query the triplestore, present basic metrics, and perform analysis upon the structure of the graph.
Currently, all events over time are being shown. An additional steop would be to add a date range picker to allow for time-span metrics and comparisons.
Summary metrics can include:
- Total sessions
- Total page views
- Bounce rate
- Avg. Time on site
The next step is to look at interesting questions that can be displayed via charting – some basics include:
- Session count over time, as a line chart
- View count over time, as a line chart
- Page paths, as a Sankey diagram
- Browser specs, as a radial diagram
Beyond that, statisical analysis can start to occur:
- What views are correlated with each other?
- What view paths are correlated with events?