Start with your decision.
Then inspect the data.
Download expanded synthetic clickstream events, AI search signals, and domain trend packages. Use the buyer map to find the fields, metrics, and SQL pattern that fit your team.
Three universal technical starting points
Each package now includes broader synthetic category and market examples. The structure is designed for warehouse and methodology review, not as evidence of current BGraph scale, coverage, or availability.
Clickstream events feed
A synthetic event-level feed aligned to the supplied raw field specification, including hashed randomized user IDs, UTC timestamps, geography, referrer and target URLs, user agent, and HTTP code.
- No real users, browsing records, URLs, cities, or postal codes
- Pseudonymous event-level delivery requires a separate rights and privacy review
- Production fields may be removed, generalized, aggregated, or delayed
AI search signals
An aggregated market view of AI discovery by broad platform class, topic, intent, source domain, destination domain, referral activity, cohort measure, and time to visit.
- No raw prompt text, private conversation text, or person-level profile
- Counts, rates, taxonomies, and privacy tiers are illustrative
- Small cohorts, sensitive topics, and disallowed uses must be excluded
Domain and category trends
An aggregated sample for indexed domain movement, category rank, cohort reach, referral share, AI referral share, channel mix, and week-over-week change.
- Index values are not traffic or panel estimates
- Shares are illustrative and may omit unlisted channels
- Taxonomies, normalization, cohorts, and rank scope require documentation
Choose by the decision your team needs to make
These profiles map the three samples to common evaluation questions. They are not separate products or coverage promises.
Market intelligence and consulting
Benchmark category leaders, challengers, and channel shifts.
Index, rank, referral shareInvestment and due diligence
Compare synthetic momentum, demand paths, and competitive concentration.
Change, cohort index, transitionsRetail and ecommerce
Understand discovery-to-product journeys and referral mix.
Paths, CTR, time to visitTravel and hospitality
Measure destination discovery, research, and booking transitions.
Intent, referrals, pathsSaaS and product teams
Track research, comparison, validation, and vendor visits.
Intent, CTR, category rankAI search and GEO teams
Monitor AI discovery classes, topics, destinations, and downstream visits.
Topic, platform class, referralsMedia and publishers
Evaluate research destinations, audience movement, and AI referral mix.
AI share, search share, cohort indexAdvertising agencies and adtech
Compare synthetic channel shares and category demand paths.
Referral, search, direct, pathsConsumer brands
See how category research moves toward product and comparison pages.
Intent, destination, time to visitData platforms and ML teams
Validate schemas, ingestion patterns, feature candidates, and joins.
Fields, windows, privacy tiersCybersecurity and risk research
Study vendor discovery without person-level profiling.
Topic, intent, visits indexAcademic and policy research
Evaluate aggregated behavioral methods, taxonomy, and reproducibility.
Cohort indices, channel sharesMore category shapes for integration testing
Twelve category and market combinations make the same schemas useful across consumer, B2B, media, finance, travel, and technology evaluations.
All category names, markets, domains, paths, counts, rates, and indices are fabricated. Inclusion here does not mean that BGraph currently covers or sells that category or market.
Fields carried into the event sample
The clickstream package follows the supplied raw-data field list. Names are normalized to snake_case for warehouse use. The public CSV is convenient for evaluation, while an approved raw delivery can use partitioned TSV with gzip or Parquet.
- Raw format
- TSV
- Archive
- Gzip
- Folders
- Year / Month / Day
- Line endings
- Linux LF
- Reference file size
- Up to 1000 MB
UserIduser_id_hashHashed randomized identifier
EventDateevent_dateUTC event and partition date
EventTimestampevent_timestampUTC time with milliseconds
CountryCodecountry_codeISO country code
CitycityOptional city label
PostalCodepostal_codeOptional postal code
ReferrerUrlreferrer_urlReferring URL
TargetUrltarget_urlVisited URL
UserAgentuser_agentRetrieval and interaction agent
HttpCodehttp_codeOptional HTTP response code
Inspect how the files can be used
The downloadable SQL files use DuckDB syntax and Parquet paths. Adapt table paths and field policies to the approved delivery.
Clickstream events feed
Download SQL ↓WITH ordered_events AS (
SELECT user_id_hash, event_timestamp, target_url,
LAG(target_url) OVER (
PARTITION BY user_id_hash ORDER BY event_timestamp
) AS previous_target_url
FROM read_parquet('clickstream-events/*.parquet')
)
SELECT previous_target_url, target_url, COUNT(*) AS transitions
FROM ordered_events
WHERE previous_target_url IS NOT NULL
GROUP BY 1, 2
ORDER BY transitions DESC;AI search signals
Download SQL ↓SELECT market, query_topic, destination_domain,
SUM(referral_visits) AS referral_visits,
SUM(referral_visits * click_through_rate)
/ SUM(referral_visits) AS weighted_ctr
FROM read_parquet('ai-search-signals/*.parquet')
GROUP BY 1, 2, 3
ORDER BY referral_visits DESC;Domain and category trends
Download SQL ↓SELECT market, category, domain, visits_index,
week_over_week_change, ai_referral_share, rank_in_category
FROM read_parquet('domain-category-trends/*.parquet')
WHERE period_start = DATE '2026-07-13'
ORDER BY market, category, rank_in_category;Every value is fabricated. These packages show structure and analysis patterns only. They do not imply current users, events, traffic, countries, history, panel size, market coverage, or commercial availability.
Request a sample for your market ↗