Two incomplete views of the same journey
On-domain behavior is precise. You know which pages were viewed, in what order, and how recently. Its weakness is scope: it only sees activity on properties you control. Off-domain intent adds breadth by showing research themes across a wider set of content sources, but it can lack the product-specific context of your own site.
Used together, the two views answer different questions. Website behavior explains how a person engages with your offer. Off-domain research indicates whether the broader problem is active in their world.
Combine signals around a person and topic
A useful model needs a stable identity key, a topic taxonomy, and a common time window. Normalize website pages and open-web research into a manageable set of buying themes rather than hundreds of isolated events.
- Resolve the on-domain visitor to an approved profile.
- Map page behavior to product and problem topics.
- Match recent off-domain research to the same person and topic family.
- Apply fit and suppression rules before scoring.
Score intent with reasons, not mystery
Start with four dimensions: topic relevance, signal recency, engagement depth, and person or company fit. Keep the model inspectable so operators can see why an audience member was selected.
Weight recency deliberately
Research interest decays. A strong signal from this week should usually matter more than a broad signal from several months ago. Decay rates can vary by buying cycle, but they should be explicit.
Reward agreement across sources
When on-domain and off-domain signals point to the same topic, increase confidence. When they conflict, preserve both facts rather than forcing a false conclusion.
Separate fit from intent
A person can be highly active but poorly suited to the offer, or an ideal fit with weak current interest. Keep those dimensions visible so teams can choose the right response.
Add guardrails before activation
Intent is a prioritization signal, not proof of an individual decision. Outreach should use the context to become more relevant without exposing tracking details or making uncomfortable assumptions.
- Apply source consent, jurisdiction, and suppression requirements.
- Set minimum evidence thresholds before sales activation.
- Use broad problem context in messaging, not a list of observed pages.
- Review false positives and stale signals on a fixed cadence.
The strongest intent program is not the one with the most events. It is the one that consistently turns explainable signal combinations into better-timed, more relevant action.