Sessions

How Cohortum groups events into sessions, how session boundaries form, and how the Session range control picks a user's Nth session.

A session is a single visit — everything one user does before they step away. Cohortum groups your events into sessions so you can look at behavior one visit at a time, instead of as one endless stream of clicks.

What a session is

Think of a session as a sitting. A user opens your app, watches a video, clicks through a few pages, then leaves. Those events form one session. When the same user comes back the next morning, that starts a new session.

Sessions give your data structure. Instead of asking "what does this user do across all of history," you can ask "what happens inside one visit" — where it starts, how far it gets, and where it ends.

They also keep comparisons honest. Because sessions are numbered per user, you can line up like with like — one user's first visit against another user's first visit — instead of comparing someone's 1st session with someone else's 10th. A newcomer's first visit and a regular's tenth behave nothing alike, so mixing them muddies every analysis.

How sessions are formed

You decide how sessions are drawn when you upload a source, on the final Sessions step of the new-source wizard (Source → Upload → Mapping → Sessions). There are two ways to draw them — and whenever you can supply your own session ids, we recommend it: your logs hold each user's full history, so the boundaries and their numbering match your own definition exactly.

Pick a session column whenever something upstream — an analytics SDK, a data-warehouse export, or your own preprocessing — can stamp each event with a session id; that's the most accurate option. Reach for auto-generate only when your data has no session concept of its own, which is the case for most raw clickstream logs.

The Session range control

Every analysis has a left panel with a Source block. Below the date range, you'll find a Session btwn N and M range control.

The Session range picks which of a user's sessions to include, counted in order — their 1st session, 2nd session, and so on. It is not a filter on how many sessions a user has.

Sessions are numbered from 1. Session btwn N and M includes every session whose number falls between N and M, both ends included — so setting N to 0 means "start from the very first session." For sessions 3 through 5, set the range to Session btwn 3 and 5.

Each user contributes only the sessions inside the range. A user with fewer sessions than N contributes nothing. A user with more than M keeps their in-range sessions and drops the later ones.

Example: a user's first session

Set the range to Session btwn 0 and 1.

• ✅ This includes each user's first session — the events from their earliest visit.
• ❌ It does not mean "users who have only 1 session."

A user with 20 lifetime sessions still contributes, but only their 1st session's events enter the analysis. Widen the range to Session btwn 0 and 5 to include each user's first five sessions.

This makes the control ideal for onboarding questions. Set it to the first session or two to see how brand-new visits behave — where users land, what they click, where they drop off — without later, more expert visits diluting the picture.

Analysis levels and sessions

Cohortum runs an analysis at one of three levels (shown as the Mode on Home and the Level column in Saved). Sessions apply at two of them, and the Session range control and session formation work the same way at both:

• Events Analysis works on your raw events. When you build a Steps chart, a Journey, a Transitions graph, or Clusters here, each node is a single event like "Video clicked" or "Page pagination clicked". This is where you'll spend most of your time.
• Sessions Analysis uses the same four tabs — Steps, Journey, Transitions, and Clusters — but works on session types instead of individual events. A session type is a recognizable pattern of activity inside a session, so the visualizations show how kinds of visits unfold. Since session types are derived, this level adds a Mapping button in the Sessions panel that opens the Edit session types editor.

Session numbering means the same thing at both levels: "session 3" is a user's third session whether you're reasoning over raw events or over session types.

The third level, What-If, lets you test whether users reach a goal, built on the session transition graph. It shares the same Source, date range, Session range, and filter controls as the other levels — so sessions apply here too — and lets you simulate how improving a transition moves users toward a target event.

Session types and the mapping editor

At the Sessions level, the four tabs are built on session types rather than raw events. A session type is a recognizable kind of visit — a pattern of what a session was about, rolled up from its events rather than a single click. Cohortum assigns every session a type automatically by grouping similar sessions, and the left panel lists those types (with user counts) in place of the Events checklist.

To shape that mapping, click Mapping in the left panel's Sessions section. It opens the Edit session types editor, whose golden rule sits right at the top: order matters — when more than one type fits a session, the one higher in the list wins.

In the editor you can:

• Set how many types there are — the Number of session types stepper (− / +) controls how granular the automatic grouping is; click Recompute to regroup at the new count.
• Add custom types — + Add custom session type defines a type by rule (Select condition: sessions that performed an event, or whose user had a property), so you can carve out exactly the visits you care about — say, "sessions that reached Checkout."
• Reorder by dragging the handle on each row. Order is priority: a session takes the first type it matches; whatever is left over falls through to a built-in Others type.
• Rename and recolor each type, and read the event patterns that define it — shown as chips with how prevalent each pattern is — next to its user count and share.
• Toggle a type off to drop it from the analysis without deleting its definition.

Click Save to apply the mapping. Every tab — Steps, Journey, Transitions, Clusters — then speaks in your session types instead of raw events. (Recompute only re-runs the automatic grouping after you change the number of types.)

Related