What clickstream data can represent
A clickstream is useful when events can be connected in time and interpreted inside a defined market. The unit may be an event, a transition, a session pattern, a cohort, or an aggregate journey rather than an identified person.
Useful fields depend on source rights and the intended decision. Common categories include time buckets, domains, channels, product or content categories, journey stages, aggregate cohort references, and derived transition measures.
- Discovery sources
- Domain and category visits
- Channel transitions
- Research depth
- Repeat activity
- Aggregated outcomes
Event feeds and journey aggregates answer different questions
Event-level products preserve more sequence detail and support custom modeling. They also require more engineering, documentation, access control, and quality review.
Journey aggregates reduce complexity by summarizing paths, overlap, timing, and transitions for approved groups. They are often easier to use for market research, dashboards, and executive analysis.
- Use event feeds for custom sequence and graph models
- Use aggregates for interpretable market patterns
- Confirm the minimum useful granularity during scoping
Coverage quality matters more than a headline total
A large global number does not prove that a specific country, category, device class, or period is represented well enough for a decision. Coverage should be evaluated for the requested universe.
Quality review should consider missingness, duplication, source changes, stability, small cohorts, category normalization, and possible selection bias. Output documentation should make these limitations visible.
Privacy controls belong in the product definition
Aggregation and pseudonymisation can reduce risk, but neither word creates a universal legal conclusion. The source, fields, joins, access model, geography, buyer purpose, and retention all matter.
BGraph does not position standard clickstream outputs for re-identification, individual targeting, unlawful surveillance, or unsupported high-impact decisions.
- Purpose limitation
- Field minimization
- Cohort thresholds
- Suppression
- Buyer restrictions
- Retention controls