Overview
Feature: Career Path Explorer · Owner: Pushyami Shandilya · Status: Alpha
One-liner: A dual-mode tool — Path Analytics and People Search — that lets LinkedIn members explore career transition patterns and find the real people who made them.
LinkedIn lets you search by current role, company, and location. It does not let you search by career path — "show me people who were Data Scientists and are now PMs." That gap is what this feature closes.
Problem
For career transitioners
Someone considering a DS → PM pivot has no way to find peers who made that exact move. Targeted networking — the highest-value job search activity — is entirely manual and serendipitous.
For LinkedIn
Career exploration is high-intent but underserved. Members actively planning transitions are the most likely to upgrade to Premium. There is no dedicated surface for this behavior today.
Goals & Success Metrics
| Goal | Metric | Target |
|---|---|---|
| Drive path-filtered searches | Searches with ≥2 path nodes set | 40% of sessions |
| Increase connection requests | Connect CTR from results | >8% (vs. 3% baseline) |
| Premium conversion | Upgrade rate from drawer upsell | 2.5% of drawer opens |
| Session depth | Avg. cards opened per session | ≥3 |
| Return usage | 7-day retention | 30% |
Target Users
Primary — Career transitioners
Mid-level professionals (3–8 yrs) actively considering a role change. High intent, low confidence in path viability. Want social proof and warm intros.
Secondary — Recruiters
Want to find candidates with non-linear backgrounds. Current search doesn't support trajectory filtering.
Tertiary — Explorers
Early-career members browsing possible futures. Lower intent but high engagement — drive discovery metrics.
Feature Spec
Mode 1 — People Explorer
| Component | Description | Priority |
|---|---|---|
| Path builder | 2–4 node role selector. Add/remove steps, wildcard support. | P0 |
| Insight strip | Match count, median transition time, salary change, 2nd-degree count. | P0 |
| Profile cards | Name, headline, trajectory with matched nodes highlighted, Connect/Message CTAs. | P0 |
| Profile drawer | Full timeline, skills, education, transition stats, "Request Chat" CTA, Premium upsell. | P0 |
| Filters | Industry, company size, time-to-transition, location, network degree, education. | P1 |
| Sort | Relevance, fastest transition, most connections, most recent. | P1 |
| Reverse path toggle | "Who was X before becoming Y?" direction flip. | P1 |
| Empty state | Zero-results screen with path-broadening suggestions. | P1 |
| Saved paths | Bookmark a search. Surfaces in My Network tab. | P2 |
Mode 2 — Paths Analytics
| Component | Description | Priority |
|---|---|---|
| Force graph | Role nodes, weighted edges, Sankey side panel on click. | P1 |
| River / alluvial view | Cohort flow over time. Shareable as a card. | P2 |
| "Find people" CTA | Click graph edge → People mode pre-filled with that path. | P1 |
UX Flow
Primary
- User lands on Career Path Explorer (entry: My Network tab, Jobs tab, profile sidebar)
- Selects Step 1 ("Data Scientist") → match count updates live
- Selects Step 2 ("Product Manager") → results load
- Clicks a profile card → drawer slides in
- Clicks "Request Chat ☕" → InMail compose opens pre-populated
From Paths mode
- User explores force graph, clicks DS → PM edge
- Side panel shows transition stats
- Clicks "Find people who made this move" → switches to People mode, path pre-filled
The graph edge → people results connection is the highest-value interaction. It makes both modes feel like one coherent product.
Data Model
Inputs
- Career histories: Sequence of (role, company, start, end) per member — already in LinkedIn's profile graph.
- Role taxonomy: Standardized categories mapped from freetext titles. Requires NLP normalization.
- Transition graph: Precomputed edge weights — how many members went from role A to role B.
Matching logic
- Given path [R1 → R2], return members whose history contains R1 followed by R2
- Ranking: 2nd-degree first, then recency, then path similarity score
- Wildcard: treated as don't-care in sequence match
Demo uses a synthetic dataset of 50 profiles. Matching logic runs client-side as a JS filter against profiles.json.
Privacy & Consent
- Career history used for matching only if profile is set to "Open to being found"
- Aggregate stats in Paths view are anonymized — no individual identified
- Members can opt out via Privacy Settings
- Salary data only shown for members who explicitly shared it (Premium)
Risk: Members may not expect their career history to make them discoverable. Requires clear disclosure in onboarding and a prominent opt-out.
Risks
| Risk | Likelihood | Mitigation |
|---|---|---|
| Role taxonomy quality | High | NLP classifier on LinkedIn's title graph. Manual taxonomy for top 500 titles as fallback. |
| Privacy backlash | Medium | Opt-out flow, onboarding disclosure, default to existing privacy settings. |
| Sparse results for niche paths | Medium | Empty state with broadening suggestions. Min 5 results threshold. |
| Recruiter misuse | Low | Rate-limit InMail from feature. Separate quota for LinkedIn Recruiter. |
Out of Scope (v1)
- Salary data integration
- Predictive path recommendations
- City-level geographic filtering
- Company-specific paths ("DS at Google → PM at Google")
- Mobile app
- LinkedIn Learning integration
Build Phases
| Phase | Deliverable | Status |
|---|---|---|
| 0 — Data | 50-profile synthetic dataset, role taxonomy | In Progress |
| 1 — Design | 4 visual mocks, People Explorer alpha, profile drawer | Done |
| 2 — Frontend MVP | Live filters, sort, empty state, reverse path toggle | Upcoming |
| 3 — DS layer | Real matching logic, path scoring, transition probability model | Later |
| 4 — Polish | Graph → people CTA, shareable path cards, animations | Later |