Event analysis
Event analysis is the Trend, Table, and Number tabs of the builder, which share one metric list. Step drop-off and return visits are Funnels and Retention.
ConsoleAll-platform analyticsBehaviorBehavior analysisHow to configure
Up to 12 measurements: a built-in, an event, a saved metric, or a formula. Page views, visitors, sessions, bounce rate, and engaged rate are built-ins with the same definitions as the web overview. Event aggregations are total, uniques, average per subject, active percent, sessions, first time, and property sum, average, median, 90th percentile, distinct count, min, and max. A formula is arithmetic over TOTALS, UNIQUES, AVERAGE, ACTIVE, PROPSUM, PROPAVG, or {id} for a saved metric.
Behavior is the date range (default last 28 days) and granularity (auto, hour, day, week, month). Breakdown is up to three layers, using the same dimensions as filters; the page dimension prefers the screen name on apps. Compare is previous period (optional weekday alignment), same period last year, custom, or none. Trend and number default to the previous period. Identity defaults to visitor. Several segments intersect. See Measurement × Behavior.
How to read the result
Trend shows current, previous, and change above the chart. The chart can be area, line, bar, or stacked bar, with an optional goal line. A point can go on a dashboard, become a metric, open people, or become a segment. Number keeps a single value. Table is the result, 1 to 50 rows, and says when it is truncated. Export is CSV.
Limits & definitions
Uniques follow identity: visitor, device, person (identified id, else device, else visitor), or account key. Rows without an account key are left out of the account count. Sessions fall in the bucket of their start time. Bounce is the share of sessions with one page view and no custom event. Engaged is duration of at least 10 seconds, or at least 2 pages, or a key event. Active users skip server events, names that start with a double underscore, app-end, and performance samples. The longest range is 731 days. Creating a monitor from a point only stores a draft when that action is not connected yet.
Property analysis
The property does not have to be registered as a breakdown dimension first. Numeric buckets you can chart today are Property value on Distribution; splitting a metric by a registered dimension is the breakdown on Event analysis.
ConsoleAll-platform analyticsBehaviorBehavior analysisHow to configure
The builder lists Properties, but it has no property-name field and the chart does not know this result, so the panel is empty. The server also rejects the run when the property name is blank.
The server requires a property name, 1 to 60 characters. An optional event name limits the rows; empty means every event in the range. The value list defaults to 20 and caps at 100. Time, filters, segments, and identity apply; see Measurement × Behavior. No breakdown and no compare period. Identity defaults to person, and the people count uses that key.
How to read the result
The payload has events that carry the property, events in range, coverage (the first divided by the second), and the number of distinct values. When at least 80% of non-empty values parse as numbers, it also returns min, max, average, median, 90th percentile, and sum, from at most the latest 50,000 rows. The value table is events, subjects, and share of events that have the property. The trend keeps only the 5 values with the most events and plots each one’s event count at the builder’s granularity. None of this is drawn in the builder, and CSV export does not include it.
Limits & definitions
Nulls stay out of the value table but stay in the event total, so coverage can be under 100%. The numeric check requires the whole text to look like a number; mixed text produces no numeric row. Distribution’s property mode is different: it requires an event and only buckets numbers.