1. Start with data you control.
A user-provided LinkedIn archive can include connections, your position history, message metadata, recommendations, endorsements, invitations, job activity and other professional context. RelationshipsIQ is designed to use approved source data without requiring your LinkedIn password or making scraping the core workflow.
2. Reconstruct people before you rank them.
The same person can appear across connection records, messages, recommendations and other files. Identity resolution must combine those appearances conservatively before relationship strength, role inference or goal relevance can mean anything.
3. Separate relationship evidence from the current objective.
Relationship strength can be built from evidence such as interaction volume, recency, reciprocity, duration, recommendations and shared context. Goal fit is a different question. The same strong relationship can matter greatly for one objective and barely at all for another.
4. Use time correctly.
One archive already contains historical message timestamps, which can support Relationship Weather such as strengthening, cooling or dormant strong ties. A second export adds something different: snapshot intelligence. Comparing two archives can show company changes, title changes, founder transitions, new connections and shifts in goal relevance.
5. Ask network-level questions, not just person-level questions.
Once people and relationships are normalized, the archive can support company concentration, organizational coverage, former-colleague dispersion, Network Runway, Relationship Capital and other structural findings. Those answers reveal whether the network has credible capacity—not merely whether matching contacts exist.
What a LinkedIn export cannot tell you
This boundary is essential. A LinkedIn export is effectively a star graph: it describes your connection to each person, not the connections between those people. It cannot prove that Mark knows Sarah, report a true mutual-connection count, or support an honest “82% introduction probability.” It also contains only the current title recorded at download time for a connection; it does not contain that person's full career timeline.
It can identify your direct relationship to a relevant person, your relationship to people inside a target employer, role and company coverage, bridges created by the domains a person spans, and changes visible between snapshots. Where the archive cannot prove a second-degree relationship, the product should say so.
6. Keep private content out of the answer when metadata is enough.
The core product is designed to discard private message body content during parsing and retain only the relationship metadata needed for analysis. Sensitive files such as logins, security challenges, phone numbers, identity verification and ad data do not belong in the relationship graph.
7. Let the user correct the archive.
An export cannot know every client, mentor, friend or advisor relationship. User verification adds what the archive cannot infer and outranks machine inference in future ranking.