What stays in your browser and what normal site operation records.

What may be stored

Individual sanitized events and limited metadata help AscendLab understand tool discovery, search behavior, saved tools, related-link clicks, workflow discovery, errors, coarse performance signals, and aggregate usage trends. Metadata includes page paths without query strings or fragments, browser and operating-system families, device class, viewport dimensions, network hints, country-level location, timestamps, and random identifiers. Your analytics choice, its version, and its recorded time accompany accepted first-party events.

• route
• locale
• surface
• tool_slug
• tool_category
• category
• target
• action
• experiment_id
• variant
• result_count
• result_position
• saved_count
• recent_count
• query_length
• query_token_count
• token_count
• has_query
• no_result
• source
• device_type
• browser_family
• browser
• os
• network_type
• save_data
• event_id
• session_id
• anonymous_user_id
• timestamp
• page_path
• source_page
• viewport_width
• viewport_height
• web_vital
• metric_value
• metric_delta
• rating
• navigation_type
• action_name
• result_success
• input_type
• duration_ms
• file_size_mb
• file_type
• file_size_bucket
• output_size_kb
• settings_preset
• wasm_load_ms
• error_code
• error_stage
• country
• language
• os_family
• traffic_source
• user_agent_family
• anonymous_session_hash
• returning_visitor
• session_age_seconds
• session_id_hash
• visitor_age_days
• visitor_id_hash
• release_id
• release_sha
• build_id

What is not stored

Product analytics are designed not to store public tool input, generated output, raw search text, uploaded file names, uploaded file contents, clipboard contents, full IP addresses, full user-agent strings, full referrer URLs, or precise location.

• raw tool input
• raw uploaded file content
• uploaded file name
• full clipboard content
• exact geolocation
• full IP as product event metadata
• full user-agent string
• full referrer URL
• canvas or font fingerprint
• cross-site advertising identifier
• personal pasted text
• raw sensitive query
• generated result content

Search metadata

Search events use query length, token count, result count, and no-result flags so AscendLab can improve search without retaining the raw query. Referrers are reduced to categories such as direct, search, social, or internal.

Random visitor and session IDs

After acceptance, product analytics use first-party random identifiers stored in this browser to estimate unique visitors, sessions, and returning visits. These identifiers can link individual events, so the records are pseudonymous rather than fully anonymous. The IDs are not derived from IP address, user-agent fingerprinting, tool input, or cross-site tracking. A visitor ID is reused for up to 180 days from creation; a session ID is reused until 30 minutes of inactivity. Expired IDs are replaced on the next analytics event, rather than automatically erased from browser storage at those deadlines.

Human and bot-like traffic

Admin analytics can separate likely human traffic from bot-like traffic using coarse user-agent family and device hints. AscendLab does not store full user-agent strings for this product-event view.

QA and internal events

Internal QA events may be retained for release validation, but admin analytics separate those events from the default real-traffic overview. Ordinary localhost and non-production product events are skipped unless they are explicitly marked for QA.

Admin exports

Admin CSV exports use the same sanitized product-event fields shown in the dashboard. Some fields may be expanded into columns for filtering, but exports are intended for operational review, not session replay or raw visitor profiling.

Retention

The maintenance task defaults to deleting product events, tool-job summaries, error records, performance samples, and product feedback older than 180 days; the cutoff is configurable. This is not an enforced automatic deletion deadline, and records can remain until maintenance runs. Exports, backups, provider records, and security logs are separate and are not covered by this cutoff.

Your analytics choice

First-party analytics requires acceptance in Privacy settings; Google Analytics, currently disabled, also requires acceptance if enabled. A refusal saved under the previous notice remains effective; a previous acceptance requires a new choice. The versioned choice and its date remain in browser storage until changed or cleared, with no automatic expiry. If this browser cannot save your choice, first-party analytics stays off. Reject analytics opts out of both first-party analytics and Cloudflare measurement, removes our stored analytics IDs and accessible Google Analytics cookies, and reloads the page to stop an initialized Cloudflare probe. Requests already sent or in flight cannot be recalled, and records already received are not erased. Tools remain available. This choice does not authorize advertising or control necessary delivery and security logs.

Cloudflare performance measurement

Cloudflare measurement may run before you choose, within the existing exclusion for requests connecting to Cloudflare data centers in the EEA/EU, the United Kingdom, and Switzerland. The connected data center may be in a different region from the visitor. Acceptance does not expand this collection scope. Cloudflare says the probe does not read or write browser storage; it can still send page URLs, referrer information, and performance measurements. Reject analytics opts out, with the reload boundary described above.

Cloudflare documents seven days of unsampled beacon data, then aggregation and sampling for longer-term storage. Access to six months of reports is not a deletion guarantee. Privacy settings do not delete provider records. See the Cloudflare RUM documentation and retention FAQ.