01
Capability boundary
Profile viewers are absent from the DewEngine catalog
The target LinkedIn resources include profiles, relations, search, posts, jobs, chats, messages, companies, and invitations. They do not include profile-view observations or a viewer-list method. The connector is also unimplemented and decision-gated. DewEngine cannot retrieve viewers, and a team must not scrape pages or replay private requests to manufacture that capability.
- Keep the feature absent from SDKs and product UI
- Require an explicit provider contract before adding a viewer resource
02
Purpose review
Viewing activity is not automatic sales intent
Even if an authorized product exposed some viewing context, a visit could have many meanings and may be partial or anonymized. Do not label a viewer a lead, infer interest, or enroll outreach automatically. Define a narrow user-facing purpose, such as showing the account owner their own authorized activity context, and preserve uncertainty in every downstream interpretation.
- Avoid intent scores based only on a view
- Keep any observation separate from contact permission
03
Observation schema
Model time, subject, visibility, and confidence
A viewer observation would need the viewed account, observing account, viewer identity only if actually returned, observation window, source product, visibility level, and retrieval time. Anonymous or aggregate activity cannot be reverse-resolved into a person. Counts and ordering are mutable observations, not a durable relation graph. Missing entries do not prove that no one viewed the profile.
- Represent anonymous and identified observations distinctly
- Do not combine separate partial signals to deanonymize a viewer
04
Retention
Prefer short-lived review over permanent tracking
Viewer activity can become sensitive behavioral data. Keep the minimum fields for the stated feature, restrict them to the account owner or authorized role, and use a short documented retention period. Avoid broad exports, cross-customer identity graphs, and model training. Deletion and disconnect should stop future collection and remove cached or derived viewer signals according to policy.
- Audit bulk access to behavioral observations
- Keep viewer data out of generic CRM enrichment by default
05
User experience
Show uncertainty and avoid coercive automation
If such a feature were ever authorized, the interface should state the observation window, source, incomplete coverage, and whether identity is unavailable. It should not encourage immediate automated contact. A user may choose to research a known existing relationship through normal customer records, but the viewing signal alone cannot justify a message or invitation.
- Use neutral language instead of hot-lead labels
- Require normal suppression and approval for any later action
06
Release gate
Demand viewer-specific authorization and tests
A future capability would need provider and legal approval, exact eligible products, visibility semantics, anonymous handling, pagination or aggregation behavior, freshness, throttling, restriction, tenant isolation, retention, export, correction, deletion, and consented conformance. No adjacent profile or search evidence can substitute. Until a manifest contains this proven resource, the answer remains unavailable.
- Test that anonymous activity cannot be reidentified
- Remove the feature if evidence or authorization expires