Finding Sessions To Watch

Most people open session replay with a goal (“why are signups dropping?”) but no idea which of thousands of recordings to watch. A raw, unfiltered list is the worst possible answer — it buries the useful sessions in noise. Your job is to turn their intent into a focused filter, return a handful of high-signal recordings, and offer to dig into one.

The starting points below are the same ones the product surfaces as “filter templates” — they encode the jobs people actually use replay for. Treat them as a menu, not a script.

The one rule

Never dump an unfiltered recording list. Always either (a) apply a goal-based filter, or (b) sort by a signal (activity, errors) so the first few rows are worth a click. If the user’s goal is unclear, ask one short question or offer the menu before querying.

Available tools

Hand off to the investigating-replay skill once the user picks a recording to understand in depth.

Workflow

1. Pin down the goal

Map the request to one of the starting points below. If it’s vague (“show me something interesting”), offer 3-4 options rather than guessing, or default to most active sessions (high signal, no setup).

2. Discover before you filter

Event names and URLs vary per project — never assume $pageview paths, a signup_completed event, or a person property exists. Confirm with read-data-schema (event_properties, event_property_values, entity_property_values) before putting a value in a filter. If the needed event/property doesn’t exist, say so and suggest the closest available signal.

3. Run a minimal query

Call query-session-recordings-list with only the filters that serve the goal. Recommended settings:

• set filter_test_accounts: true (the tool defaults to false) to exclude internal users, unless the user is debugging their own session.
• date_from of -7d to -30d for goal-based searches; -3d for “recent”.
• A deliberate order — activity_score for “interesting”, console_error_count for “broken”, start_time for “recent”.
• limit: 10 — you want a shortlist, not a dump.

4. Triage and present

Don’t relay raw rows. Pick the 3-5 most promising and say why each is worth watching (long active duration, many errors, reached the key page, high activity score). Deep-link each as {posthog_base_url}/replay/{id} — never /replay/home?sessionRecordingId={id}. Note total matches so the user knows how much is behind the shortlist.

5. Offer the next step

• “Want me to walk through one?” → investigating-replay.
• “Want to keep watching these?” → save it as a saved filter view with session-recording-playlist-create (type: 'filters' — a filter view, not a 'collection', which is for manually curated recordings and can’t carry filters).

Starting points → filters

Two filter shapes cover almost everything:

• Reached a page → recording metric visited_page ({ "type": "recording", "key": "visited_page", "operator": "icontains", "value": "/pricing" }).
• Did a specific event (signup, search, rageclick, used a feature) → there is no event-name filter on the recordings query, so first collect session IDs with execute-sql, then pass them as session_ids (see the two-step pattern below).

Two-step pattern: “sessions where event X happened”

The recordings query filters by event properties, not event names. To find sessions that contain a particular event, collect the session IDs first:

posthog:execute-sql
SELECT $session_id
FROM events
WHERE event = '$rageclick'          -- or your signup/search/feature event (confirm via read-data-schema)
    AND timestamp > now() - INTERVAL 7 DAY
    AND $session_id != ''
GROUP BY $session_id
ORDER BY max(timestamp) DESC

Then fetch those recordings (some session IDs won’t have a recording — that’s expected). Pass the same date_from window as the SQL step — with only session_ids, the query falls back to its -3d default and would drop sessions whose event was older than that:

posthog:query-session-recordings-list
{ "date_from": "-7d", "session_ids": ["<id1>", "<id2>", "..."] }

Worked example

User: “Why are people bouncing on our pricing page? Show me some sessions.”

1. Goal = pricing-page friction → visited_page approach.
2. read-data-schema (event_property_values for $pathname) to confirm the path is /pricing.
3. Query:

posthog:query-session-recordings-list
{
  "date_from": "-14d",
  "filter_test_accounts": true,
  "order": "activity_score",
  "limit": 10,
  "properties": [
    { "type": "recording", "key": "visited_page", "operator": "icontains", "value"

4. Present the 3-5 most active, each as {base}/replay/{id}, noting which lingered or hit errors.
5. Offer to investigate the most promising one (investigating-replay) or save it as a saved filter view (type: 'filters').

Tips

• Prefer one good filter over many — over-filtering returns nothing and reads as “no data”.
• If a query returns zero recordings, widen the date range or loosen the filter before concluding there’s nothing to watch; if it’s still empty, recordings may not be captured for that flow (point the user to diagnosing-missing-recordings).
• activity_score is a solid default proxy for “worth watching” when there’s no sharper signal — but it rewards raw interaction volume, so prefer a goal-based filter (errors, a key page) when you have one.
• Keep the shortlist short. The value is in choosing for the user, not handing back the haystack.