B2B buyers move between work and personal devices, shared browsers, and colleagues’ links. Measurement may connect some interactions when a person is authenticated and the organization has a valid basis, but gaps remain.
Distinguish observed and modeled links
A deterministic connection uses a stable identifier available in an authorized context, such as an authenticated account. A modeled estimate uses patterns to infer likely connections. Reports should distinguish observed links from estimates where possible.
A device cookie or IP address is not a reliable person identifier by itself. Shared networks, privacy protections, consent choices, and device changes can make apparent cross-device paths incomplete.
Respect consent and data minimization
Use only identifiers necessary for an authorized purpose and honor recorded preferences. Do not reconstruct identities from sensitive combinations of device, location, and behavior signals.
Document where identifiers are collected, how they are protected, and when they are deleted. Cross-device measurement should not become a reason to retain personal data indefinitely or bypass a person’s choice.
Report gaps as part of the result
Show the share of sessions or conversions with an eligible identity link when available, and explain which platforms contribute modeled estimates. Do not present an incomplete journey as the complete path.
The same conversion can have different platform paths depending on login status, attribution settings, and consent. Reconcile channel performance with CRM outcomes and customer research.
Design decisions that do not require perfect identity
Many planning decisions can use aggregate trends, experiments, and cohorts without joining every device. Choose a method that matches the decision’s precision needs and privacy constraints.
Preserve uncertainty in reporting. For related technical context, see our guide to cross-domain B2B attribution.
Related reading: cross-domain B2B attribution.
Practical checklist
- Name the decision and the evidence it requires.
- Record the definitions, assumptions, and ownership that affect interpretation.
- Revisit the method when the underlying process or data changes.
How did this article land?
Choose one reaction. You can change it anytime.