Analyze sessions in Metric Explorer

Session metrics in Metric Explorer measure visits instead of individual events: how many sessions happened, how long they lasted, and what happened in them. Each session metric is its own series, with its own filters and group-bys, so one chart can show total sessions, session duration, and bounce rate side by side, next to event metrics.

A session is a period of continuous activity by one user. Amplitude groups a user's events into a session by session ID. By default, mobile SDKs start a new session after 5 minutes without events, and the Browser SDK after 30 minutes. Session metrics count only valid sessions, meaning sessions with a session ID other than -1. Events sent with a session ID of -1 are out of session and belong to no session.

Session metrics need session tracking in your project. Amplitude SDKs send session IDs automatically. If you send events through the HTTP API, include a session_id. For session tracking and custom session definitions, go to Track sessions.

Add a session metric

1. In the Metrics block, click + Add and select an event:
   • To count every session, select [Amplitude] Any Active Event.
   • To count sessions that include a specific action, select that event, for example Add to Cart.
2. Open Measured as, go to Session, and choose a measurement, for example Total Sessions.

Amplitude counts the sessions that contain the event at least once. For example, Add to Cart measured as Total Sessions counts sessions with at least one add-to-cart. For what each session measurement calculates, go to Choose the right measurement.

To compare several session measurements, add one metric for each. For example, add Total Sessions, Average Session Duration, and the [Amplitude] Bounce Rate metric to the same chart, each with its own filters.

Filter sessions by what happened in them

Session filters decide which sessions a session metric counts. To add one:

1. On the session metric, click + Filter by.
2. Choose a session filter:
   • [Amplitude] Contains Event: Sessions that contain an event a number of times. Choose the event, an operator (=, ≠, ≥, ≤, >, or <), and a count.
   • [Amplitude] Session Duration: Sessions longer or shorter than a length you set.
   • [Amplitude] First Property Value: Sessions where the first value of a property matches, for example the first [Amplitude] Page Path is /pricing.
   • [Amplitude] Last Property Value: Sessions where the last value of a property matches, for example the last page path is /checkout.
3. For [Amplitude] Contains Event, add property conditions to the event if you need them, for example Product Viewed where category is shoes.

Combine filters on one metric to describe a visit. All conditions must match:

A count of 0 with = describes sessions without an event. Segments in the Segment by block still apply, so a session counts only when its user matches the segment.

Group sessions by entry, exit, or any value

A group-by on a session metric splits sessions by a property value. Because a session can contain many values of the same property, choose which value assigns the session to a group:

1. On the session metric, click + Group-by.
2. Choose how to read the value:
   • [Amplitude] Any Property Value: The session counts under every value it contains. The sum across groups can exceed the total number of sessions.
   • [Amplitude] First Property Value: The session counts under the first non-empty value, such as the entry page or the channel that started the visit.
   • [Amplitude] Last Property Value: The session counts under the last non-empty value, such as the exit page.
3. Choose the property, for example [Amplitude] Page Path or utm_medium.
4. Choose the event that carries the property:
   • The metric's own event: Amplitude reads values only from that event. For [Amplitude] Any Property Value, this is the default.
   • [Amplitude] Any Active Event: Amplitude reads values from any active event in the session. For first and last values, this is the default.

For example, [Amplitude] Any Active Event measured as Total Sessions and grouped by the [Amplitude] First Property Value of [Amplitude] Page Path counts visits by landing page.

Bounce, entry, and exit

Amplitude defines these session outcomes with the [Amplitude] Page Viewed event and a property such as the page path:

• Bounce: A session with exactly one page view. Bounce rate is bounced sessions divided by all sessions.
• Entry: The first non-empty value of the property in a session. Entry rate for a value is sessions that entered on that value divided by all sessions.
• Exit: The last non-empty value of the property in a session. Exit rate for a value is sessions that ended on that value divided by all sessions.

To chart these rates, add the [Amplitude] Bounce Rate, [Amplitude] Entry Rate, or [Amplitude] Exit Rate metric from the metric picker. To build a rate for specific pages, combine two session metrics in a formula. For example, bounce rate by landing page:

• Metric A: [Amplitude] Page Viewed measured as Total Sessions, filtered by [Amplitude] Contains Event [Amplitude] Page Viewed = 1.
• Metric B: [Amplitude] Page Viewed measured as Total Sessions.
• Group both A and B by the [Amplitude] First Property Value of [Amplitude] Page Path.
• Formula C: %:A/B.

For a worked example of how the group-by property decides bounce, entry, and exit across several sessions, go to Use session metrics in Data Tables.

Combine session metrics with event metrics

Session metrics and event metrics can share a chart. For example, put Add to Cart measured as Event Totals next to Total Sessions to see whether add-to-carts track traffic. To relate them in one number, such as add-to-carts per session or revenue per session, write a hybrid formula. Go to Write formulas in Metric Explorer.

Date ranges with session metrics

When a chart includes a session metric, the date range can cover up to one year: 365 days, 52 weeks, 12 months, or 4 quarters. If you choose a longer range, Amplitude shortens it and shows "The date range was shortened to one year because this chart includes a session metric."