Analyzing Experiment Session Replays

This skill guides you through analyzing session recordings for experiment variants to understand behavioral differences between control and test groups.

When to use this skill

Use this skill when:

• The user asks to analyze session replays for an experiment
• The user wants to understand how users behave differently across experiment variants
• The user asks to compare user behavior between control and test variants
• The user wants qualitative insights to complement experiment metrics
• The user asks questions like “How are users behaving in my experiment?” or “Show me session replays for variant X”

Prerequisites

Before analyzing session replays:

1. The experiment must be launched (not in draft state)
2. Session replay must be enabled for the project
3. Users must have been exposed to the experiment variants
4. The experiment must have a start date

Workflow

1. Get experiment details and feature flag variants

First, retrieve the experiment information and the feature flag variants (source of truth).

Step 1a: Get experiment metadata

You can either:

• Option A: Use the experiment-get tool if you already have the experiment ID from context
• Option B: Query the experiments table via HogQL:

SELECT
    e.id,
    e.name,
    f.key AS feature_flag_key,
    e.start_date,
    e.end_date
FROM system.experiments e
JOIN system.feature_flags f ON f.id =

From the experiment data, extract:

• feature_flag_key: The feature flag controlling the experiment
• start_date and end_date: The experiment’s time range

Step 1b: Get variants from the feature flag

IMPORTANT: Always get variants from the feature flag, NOT from experiment.parameters.feature_flag_variants. The parameters can be out of sync or deprecated. The feature flag is the source of truth.

Query the feature flag to get the current variants:

SELECT filters.multivariate.variants AS variants
FROM system.feature_flags
WHERE key = '<feature_flag_key>'

Select the variants path directly — selecting the whole filters object gets truncated in results for flags with large targeting configs. Example structure: [{"key": "control", "name": "Control", "rollout_percentage": 50}, {"key": "test", ...}]

The variant key values (e.g., “control”, “test”, “variant_a”) are what you’ll use to filter session recordings.

2. Build session recording filters for each variant

For each variant in the experiment, construct recording filters that match users exposed to that variant.

Filter structure for a variant (input to query-session-recordings-list):

{
  "date_from": "<experiment.start_date>",
  "date_to": "<experiment.end_date or current time>",
  "filter_test_accounts": true,
  "properties": [
    {
      "type": "event",
      "key": "$feature/<feature_flag_key>",
      "operator": "exact",
      "value": ["<variant_key>"]

Key points:

• The $feature/<flag_key> event property records which variant the user saw — filtering on it matches recordings containing at least one event from that variant
• value is an array of variant key strings (e.g. ["control"]); for boolean flags use ["true"] or ["false"]
• Avoid the type: "flag" / flag_evaluates_to property filter for variant scoping — the recordings query accepts it but silently ignores it, returning unfiltered results (last verified 2026-06-10). If you want to try it anyway, verify it actually filters first: a query with a nonexistent flag key should return zero recordings
• Set the date range to the experiment’s start and end dates
• Enable filter_test_accounts: true to exclude test users

3. Retrieve recordings for each variant

Use the query-session-recordings-list tool with the filters constructed in step 2.

Call the tool once per variant to get recordings for each group:

• Variant “control” → recordings for control group
• Variant “test” → recordings for test variant
• Additional variants if the experiment has more than 2

The tool returns a list of recordings with metadata including:

• distinct_id — the person’s distinct ID
• recording_duration, active_seconds, inactive_seconds
• click_count, keypress_count, mouse_activity_count
• console_log_count, console_warn_count, console_error_count
• start_url — first page URL visited
• start_time / end_time, activity_score

4. Compare and analyze

Compare the recordings between variants by looking for:

Quantitative patterns:

• Session duration differences
• Activity levels (clicks, keypresses)
• Console error rates
• Bounce rates

Qualitative insights:

• User confusion or frustration indicators
• Different navigation paths
• Feature discovery patterns
• Error recovery behavior

5. Present findings

Summarize the behavioral differences between variants, highlighting:

• Total recordings per variant
• Notable behavior patterns unique to each variant
• Usability issues or friction points observed
• Recommendations based on the qualitative data

Example interaction

User: "How are users behaving in my checkout experiment?"
Agent steps:
1. Query experiment details (ID: 123, feature_flag_key: "checkout-flow-test", date range: 2025-01-01 to 2025-01-31)
2. Query feature flag "checkout-flow-test" to get variants from filters.multivariate.variants
3. Extract variant keys: "control" and "new-checkout"
4. Build filters for control variant:
   - Property filter: { type: "event", key: "$feature/checkout-flow-test", operator: "exact", value: ["control"] }
   - Date range: 2025-01-01 to 2025-01-31
5. Call query-session-recordings-list with control filters → 147 recordings found
6. Build filters for new-checkout variant and call query-session-recordings-list → 152 recordings found
7. Compare patterns:
   - Control: Average 3m 45s session duration, 12% console errors
   - New-checkout: Average 2m 30s session duration, 5% console errors
8. Present findings:
   "I analyzed session replays for your checkout experiment. The new checkout flow shows:

Important notes

Do not make assumptions:

• Always verify the experiment has recordings before analyzing
• Check that the experiment is launched (has a start_date)
• If no recordings are found, inform the user clearly

Filter construction:

• The $feature/<flag_key> event property is how you scope recordings to a variant
• One filter per variant — call the tool once per variant with its own filter
• For boolean flags, use ["true"]/["false"] as the value instead of a variant key

Error handling:

• If the experiment is in draft state, tell the user it hasn’t started yet
• If no recordings exist, suggest enabling session replay or waiting for user traffic
• If the variant count is unexpected, double-check the experiment configuration

Related tools

• query-session-recordings-list: Core tool for retrieving session recordings with filters
• experiment-get: Get experiment metadata; experiment-results-get for statistical results
• execute-sql: Query experiments table for details via HogQL