Session Paths

Session paths help your agent answer a practical growth question:

Which landing pages bring people in, where do those sessions end, and what should we improve next?

Use this when a top-pages report is too shallow, but a dashboard-style journey explorer is too much ceremony.

The report connects:

entry pages
in-session page changes
meaningful events
goal conversion
exit pages
terminal states: goal, drop_off, or truncated

Your agent can then decide whether the next useful step is a funnel, a retention check, or an experiment.

When to use paths

Ask your agent for paths when:

a page gets traffic but you do not know what visitors do next
a landing page looks good in page stats but does not convert
a key flow has drop-off and you need to know which pages people leave from
you want the next experiment to target a real journey, not a random page
you want to connect pages -> paths -> funnels -> retention -> experiments

Do not use paths for long-cycle user attribution. v1 is intentionally session-local: a goal only counts when it happens in the same session.

Copyable prompts

Use prompts like these with your AI agent:

Show me session paths for my-site with signup as the goal. Summarize the top entry pages, top exit pages, and the one path most worth improving next.

Which entry pages start the most sessions but fail to reach signup? For each one, tell me the most common exit page and the next funnel I should check.

Look at session paths for my-site with signup as the goal. Pick one high-traffic drop-off path and propose one narrow experiment to improve it.

Which content or docs entry pages lead to deeper product pages or signup? Ignore vanity traffic and focus on paths that suggest real intent.

What the report includes

Each path report is grouped by entry page.

For every entry page, the agent receives:

sessions: sessions that started on that entry page
conversions: sessions where the goal event happened in the same session
conversion_rate: converted sessions divided by sessions
exit_pages: the pages where those sessions ended, with conversion and drop-off attribution
tree: a compact journey tree of pages, meaningful events, goals, drop-offs, and truncations

The important distinction:

entry_page is where the session started
exit_pages[] shows where those sessions ended
tree[] shows what happened between entry and terminal state

Terminal states

A path ends in one of three terminal states:

goal: the requested goal event appeared in the same session
drop_off: the session ended without the goal
truncated: the session continued past the configured step cap without the goal

For drop_off and truncated, the terminal node includes exit_page so the agent can say where the session actually ended.

Bounds

Paths are intentionally bounded so they are safe for agent workflows:

Knob	Default	Allowed
`since`	`30d`	`7d`, `14d`, `30d`, `90d`
`max_steps`	`5`	`1-5`
`entry_limit`	`10`	`1-20`
`path_limit`	`5`	`1-10`
`candidate_session_cap`	`5000`	`100-10000`

Requests outside those bounds fail instead of silently expanding the workload.

The default workload is smaller than the maximum: 30d, 10 entry pages, 5,000 candidate sessions, 5 steps, and 5 branches per node.

The endpoint is also query-bounded:

it starts from candidate sessions, not an unbounded event scan
it uses at most two database read queries
it does not run per-entry-page fanout queries
it does not do user-level cross-session stitching
it does not use response caching in v1

CLI and MCP

If your agent is shell-first, it can use:

agent-analytics paths my-site --goal signup --since 30d --max-steps 5

If your agent is MCP-first, use the analytics_paths tool with the same bounded knobs.

The human-readable output should stay compact: top entry pages, exit attribution, terminal labels, and the next recommended analysis step.

How to interpret paths

Start with the entry page that has enough sessions to matter.

Then ask:

Which exit page has the most drop-offs?
Does the path reach the expected funnel step before dropping?
Is the goal missing because users never reach the next page, or because they reach it and fail there?
Should the next action be a funnel query, a retention check, or an experiment?

Examples:

If /blog/post-a drives visits that later convert, treat it as an acquisition asset, not just content traffic.
If /pricing is a common exit page, that might be normal comparison behavior.
If /signup is a common exit page without signup, that is a flow problem worth investigating.
If paths truncate often, reduce noise in event tracking or narrow the question with a smaller goal and fewer steps.

Good follow-up prompts

For the top path that drops off, build a funnel from page_view to signup and show the largest step loss.

For the entry page with the highest drop-off rate, suggest one experiment that could reduce exits without changing the whole page.

Check whether visitors who enter through this page return later. Use retention after reading the path summary.