Overview
Web analytics answers “how much traffic”; behavior analytics answers the next question: who did something, how often, where they stopped, whether they came back, and which way they moved. Web, app, mini-program, and server events are analyzed in one project; formulas live on each model’s page.
Console menu
The shell is All-platform analytics. A multi-platform project also shows Multi-platform overview.
| Group | Menu | What it is |
|---|---|---|
| Overview | Multi-platform overview, My dashboards, Platforms side by side | Dashboards reuse saved analyses; Platforms side by side splits by platform |
| Behavior | Event analysis | The list of saved analyses, filterable by model |
| Behavior | Behavior analysis | The builder; it opens on Funnels, and the tabs switch models |
| Behavior | Heatmaps | Clicks, rage clicks, scroll, and elements |
| Behavior | Session replay | Recording and playback of single sessions, off by default |
| Behavior | Metrics & SQL | The metric library; SQL is a tab in the builder, not API Playground |
Users, Business, Operations, and Data are the other four groups. Start at People & user lookup, Business goals, Monitors & alerts, and Data sources.
ConsoleAll-platform analyticsBehaviorBehavior analysisModels
| Model | Page | One line |
|---|---|---|
| Trend, table, number | Event analysis | Counts, uniques, and property aggregations |
| Funnels | Funnels | Arrival, drop-off, and time between steps |
| Retention | Retention | Who returned on a later period |
| Paths | Paths | Flow forward from a start or backward from an end |
| Distribution | Distribution | How counts, days, properties, or session length bucket |
| Interval | Interval | Time between two events |
| Lifecycle | Lifecycle | New, returning, resurrecting, dormant, plus stickiness and inactivity |
| Attribution | Attribution | How pre-conversion session touches share credit |
| Sessions | Session analysis | Session counts, duration, bounce, entry and exit |
| Properties | Property analysis | Values, coverage, and numeric stats for one property |
| Custom metrics | Custom metrics | Named metrics you can reuse |
| SQL | SQL playground | Read-only queries over four views |
| Heatmaps | Heatmaps & experience | Clicks and scroll on a page |
| Dashboards & sharing | Save, dashboards & share | Save, pin to a board, share a read-only link |
| Session replay | Session replay | Recording that is off by default, plus an outbound replay link on session details |
Shared time, filters, identity, breakdown, and compare are in Measurement × Behavior.
Where to start
Match the question first: Find the right analysis. For counts, open Behavior analysis and switch to Trend. For steps, stay on the default Funnels tab. Analysts and above can change and save a query; viewers see saved results.
Measurement × Behavior
Every analysis is one query: what you count (the measurement), which behavior you count it on (time, steps, or events), how you split it (breakdown), and what you compare it with; models do not all honor the same four pieces. Named metrics are Custom metrics, segment rules are Segments (the builder only stores segment ids), and windows and dedupe that belong to one model stay on that model’s page.
ConsoleAll-platform analyticsBehaviorBehavior analysisMeasurements, time & filters
Trend, table, and number put measurements on the left: a built-in, an event aggregation, a saved metric, or a formula, up to 12. Funnels and the later models do not edit that row. They edit steps and their own parameters.
Every model has time, identity, and filters. The default range is the last 28 days. Presets cover today, yesterday, rolling day ranges, this month, last month, this year, the last 12 months, or custom dates. Filter dimensions include page, host, referrer, UTM, geography, device, language, event name, channel, engaged, platform, and account key. Several segments are an intersection. Account identity cannot add segments.
Picking events
Every place that asks for an event (metrics, funnel steps, retention, paths, distribution, property analysis, segment conditions, virtual event rules, event goals, automation triggers, ad postbacks, and the commerce mapping, AARRR mapping, and core events in business settings) uses the same picker. Open it and type: it matches event name, display name, and description, and every space-separated word must match. The list is grouped by event category: category groups follow the order set in Manage categories, then preset and virtual events without a category, then Uncategorized. Groups in which no event has ever been sent (a freshly installed industry pack, for example) go after all the others. Tabs at the top narrow it to custom, preset (names starting with $ or __), or virtual events.
Each row reads display name · event name · count in the last 30 days. Without a display name only the event name shows. Counts refresh about every 10 minutes and are only for sorting and reference. Events hidden in event management never appear. Events that are registered but have never been sent are greyed out, placed last in their group, and marked Never seen. Click the star to favorite an event; favorites and the 5 most recently used events are pinned to the top. Favorites and recent picks live in this browser only, per project, and do not follow you to another browser.
Use the arrow keys to move, Enter to pick, and Esc to close. Where several events are allowed, such as a funnel step, Enter or a click toggles the event and the picker stays open; the picked events are OR. The few places that accept an event that has not been sent yet (metrics, live segments, event goals, automation triggers, postbacks, business settings) offer Use "your text" as the first row when nothing matches.
Who gets breakdown and compare
Breakdown exists on trend, table, and funnel. Trend and table allow three layers. A funnel uses only the first layer and keeps the 10 groups with the most entries. Other models ignore breakdown.
A compare period (previous period, same period last year, custom, or none) exists only on trend, table, and number. Trend and number default to the previous period, which can align weekdays. Behavior models have no compare period. Switching a trend that used the previous period into a table turns compare off.
Granularity (auto, hour, day, week, month) applies to trend, number, the funnel conversion trend, and the interval median trend.
Identity
| Identity | Distinct key | Default |
|---|---|---|
| Visitor | Visitor id | Trend, table, number |
| Device | Device id, else visitor | When one device should stay one subject across sessions |
| Person | Identified id, else device, else visitor | Funnels and the later models |
| Account | Account key; rows without it are not counted | Organization view; cannot open people |
A custom identity uses one event property as the subject. Session analysis reads the session table, which has no property column, so a custom identity falls back to person. The longest range is 731 days.