Redesigning LinkedIn Jobs for Early-Career Professionals
Early-career professionals are churning off LinkedIn Jobs before its identity graph has enough data about them to work
LinkedIn Jobs - LinkedIn Signal
Product Strategy & Design
Independent Case Study
July 2026
LinkedIn Jobs matches listings to identity, not just resumes. Its real asset is the graph of connections, endorsements, and signals. That asset breaks down for early-career professionals who structurally haven't had time to build it.
Experience the solution through a working prototype
Read the complete case study with all details in Notion
The Problem
Understanding who we're building for and why it matters
Early-career professionals (recent graduates and 0-2 years experience) actively job searching
Users churn off LinkedIn Jobs before the identity graph has enough data to work: 96% of 19-24 year-olds with LinkedIn accounts rarely or never use them, while entry-level postings are down 35% since 2023 and recent-grad unemployment sits at 5.7%
Three linked failures: cold-start matching (thin profiles get thin matches), access problem (features that would help most sit behind a paywall this segment can least afford), and feedback problem (applications vanish into silence)
Users apply through LinkedIn Apply, get matches that skew senior or generic, hear nothing back, and eventually abandon the platform - giving LinkedIn no data to improve their experience
Product question: What would need to be true for this to work at LinkedIn scale?
Context & Evidence
Research, data, and observations that informed the product direction
Over half of 19-24 year-olds have LinkedIn accounts, but 96% rarely or never use them - many created accounts only because a parent, university, or career center told them to. Source: Bixa Research / B2Linked study.
Unemployment among recent college graduates aged 22-27 climbed to 5.7% in late 2025, well above the 4.2% rate for all workers, with nearly 43% of new grads underemployed.
Entry-level postings in the US are down 35% since early 2023 as AI tools absorb tasks that used to be the career launchpad for new hires. LinkedIn's own Grad's Guide 2026 shows entry-level hiring down roughly 7% year-over-year.
Users who do apply report being ghosted, with no status updates and no reliable way to filter out fake or spammy postings. This is synthesized from secondary research rather than primary interviews.
Handshake uses verified institutional identity (your university) as its trust mechanism with 1,600+ institutions, but that trust signal expires at graduation.
Company reviews and salary transparency help candidates decide whether to apply, but do nothing for the silence that follows an application.
Product Insight
The key realization that shaped the product direction
LinkedIn is the only player with the raw ingredients to solve all three problems at once: the identity graph Handshake has to bootstrap, the reach Internshala doesn't have past year one, and enough data density to close the loop Naukri can't
Impact: This doesn't require a new product, just a new mode of the existing one - the way LinkedIn already treats Creator Mode as distinct with its own surfaces and incentives. The platform's core matching logic and monetization both quietly penalize early-career users when they should be supporting them.
The decision: Focus on relevance over volume — show fewer jobs, but make sure they actually fit.
Goals & Success Metrics
How we define success for users, the product, and the business
Move median time-to-status from effectively unmeasured to 14 days or less
Every application reaches one of four defined states (Viewed, Under consideration, Not moving forward with reason, Interview requested) within 14 days instead of disappearing into silence
Retain 1M+ additional early-career users annually who would otherwise churn, generating $10.8-14.4M incremental Premium revenue plus data-flywheel value for Talent Solutions
Match relevance rate (% of matches not immediately dismissed): +15pp vs control
Free-tier feature adoption among early-career users: 25% within 90 days
Median time-to-status: 14 days (from unmeasured)
90-day retention of early-career cohort: +20pp vs current
Premium conversion rate (all users): no more than 2% dip
Recruiter-reported applicant-quality score: should not decline
Time-to-fill for entry-level requisitions: should not increase
Support sentiment on rejection flow: monitored for backfire
Constraints
Realistic limitations that shaped the product decisions
LinkedIn's core matching and monetization both implicitly reward experience early-career users haven't had time to accumulate
Thin profiles produce thin matches - cold-start problem requires new matching signals beyond traditional experience
Two-sided marketplace: changes that help seekers must not damage recruiter experience or Talent Solutions revenue
No primary user interviews conducted - personas synthesized from secondary research (flagged limitation)
Must work within existing LinkedIn Jobs surface - can't require users to learn entirely new product
As someone who has built recommendation systems, I know relevance is harder than it looks. You need training data, user behavior signals, and constant tuning. LinkedIn has all of this. The question isn't technical feasibility — it's whether their business model allows it.
Options Considered
Different approaches evaluated before deciding on the product direction
Solution #1: Closed-loop application status (LinkedIn Signal)
Every application resolves to one of four states within 14 days: Viewed, Under consideration, Not moving forward (with reason), Interview requested
Feedback problem - applications no longer vanish into silence
• Requires recruiter-side bulk-triage tool
• Fixed rejection reasons may feel impersonal
• Recruiters might ignore the tool
• Legal exposure in disclosure-regulated markets
Solution #2: Recruiter bulk-triage queue
Multi-select panel allowing recruiters to apply status + reason template to hundreds of entry-level applications at once
Enables #1 without crushing recruiters with manual work
• New workflow recruiters must learn
• Risk of mass-rejection to clear queues
• Recruiter time-per-req might increase beyond +5% tolerance
Solution #3: Unbundled free visibility tier
Give early-career users free access to currently-paywalled features (Featured Applicant, low-competition filter)
Access problem - features that help most no longer behind unaffordable paywall
• Premium cannibalization risk
• Sets precedent that paywall is negotiable
• Other segments may demand similar unbundling
Solution #4: Recency-weighted matching mode
Reweight matching signals to favor trajectory and recency over absolute tenure
Cold-start problem - thin profiles get better matches
• Touches core ranking infrastructure
• High effort, uncertain outcome
• Could degrade match quality for other segments
Solution #5: Verified Early Talent badge
Partnership-based trust signal similar to Handshake's institutional verification
Cold-start credibility
• Depends on external partnerships LinkedIn doesn't fully control
• Long timeline, uncertain adoption
Strategic depth over working code
Browser extension I could ship
The goal was to demonstrate product thinking, not engineering execution. A working extension would be impressive but wouldn't show the depth of PM skills recruiters look for.
Prioritization
How decisions were made about what to build and when
RICE (Reach × Impact × Confidence / Effort)
Scored five candidate solutions where Reach is 1-10 relative scale (internal data not available), Impact is 1-3, Confidence is percentage, and Effort is person-months. Solution #1 (closed-loop status) + #2 (recruiter triage) scored highest at 2.70 and 1.68 RICE respectively, chosen as Release 1 because they have the best confidence-to-effort ratio and generate labeled data that more ambitious fixes (#4 matching) would eventually need. #3 (unbundled visibility) deprioritized to Release 2 due to Premium cannibalization risk requiring careful monitoring. #4 (recency matching) deferred to Release 3 as it's highest-effort with uncertain outcome - better sequenced after Release 1 generates training data. #5 (verified badge) moved to backlog as partnership-dependent with unclear timeline.
Scoring Table
Product Direction & Solution
The product approach chosen and how it addresses the problem
LinkedIn Signal - a closed-loop application-status system paired with a recruiter-side bulk-triage tool, so every application resolves to one of four defined states within 14 days instead of disappearing into silence. Named to reflect the core idea: early-career applicants should get a signal back, not silence.
Key Features
Four-state status field visible in Job Tracker: Viewed, Under consideration, Not moving forward (with reason), Interview requested
Recruiter bulk-action panel: multi-select applicants, apply status + reason template in one action
Auto-transition to Viewed when Hiring Assistant opens or scores an application
21-day auto-close default for untouched applications with disclosed reason
In-app and optional email notification on status change
Org-editable rejection-reason template library
User Flow
User applies via LinkedIn Apply, application appears in Job Tracker, recruiter triages using bulk panel or Hiring Assistant auto-marks as Viewed, status updates within 14 days with reason if rejected, user can take action based on clear signal instead of wondering
Design & Architecture
Job Tracker page showing four status pills and reason line for rejections
Recruiter requisition triage page with bulk-selection toolbar and status column
MVP & Roadmap
What ships first, what comes later, and what we deliberately won't build
- ✓
FR1: Four-state application status field in Job Tracker
- ✓
FR2: Recruiter bulk-action panel with multi-select
- ✓
FR3: Auto-transition to Viewed via Hiring Assistant
- ✓
Fixed rejection-reason list (5-6 options instead of editable templates)
- →
FR4: 21-day auto-close default (legal review required)
- →
FR5: In-app and email notifications
- →
FR6: Org-editable rejection-reason templates
- →
Application Readiness nudge (pre-apply profile-completeness prompt)
- →
Match Explanation feature (fold into Release 3 with recency matching)
- ✕
Explainable Job Fit Score as standalone release - explaining a bad match exposes weakness rather than building trust; better bundled with improved matching in Release 3
- ✕
Applicant-level A/B testing - creates contaminated results; requisition-level cluster randomization required despite higher sample-size need
- ✕
Platform-wide launch - scoped to entry-level requisitions and early-career applicants initially
Automated apply-to-all feature
Would increase application volume but decrease quality. Goes against the core insight of relevance over volume.
Endless scroll feed
Infinite scroll optimizes for engagement, not outcomes. Wanted to respect user time with a defined daily set.
Measurement & Experimentation
How we validate the product decisions and measure impact
If entry-level requisitions get closed-loop status instead of current silent default, applicants will show higher 30-day platform re-engagement and self-reported trust, without measurable increase in recruiter time-per-requisition
Cluster-randomized controlled trial where randomization unit is the entry-level requisition (not applicant). Treatment: full FR1+2+3+MVP reason list. Control: current experience. Duration: 8 weeks minimum. Sample: 800-1,200 requisitions per arm, stratified by employer size and industry.
- ✓
Median time-to-status moves to 14 days
- ✓
30-day Early-Career Activation Rate (ECAR) increases
- ✓
Applicant trust survey scores improve
- ✓
Recruiter time-per-req stays within +5% tolerance
Early-career applicants to treatment requisitions receive status within 14 days, show higher return-to-platform rates at 30 days, and report higher trust scores versus control group still experiencing silence
Phase 0: Employer opt-out with default-on for Hiring Assistant users. Phase 1: Post-experiment ramp from 25% to 60% to 100% over 3 weeks if guardrails hold. Phase 2: US/English markets first, EU second (legal review required), India/high-volume markets third.
Learnings
What I learned, what surprised me, and what I'd do differently
What I Learned
- •
The highest-confidence fix isn't always the flashiest - closing a feedback loop scored better than AI matching overhaul
- •
Two-sided marketplace constraints are real: every change helping seekers had to be stress-tested against recruiter impact
- •
Data gaps should be flagged honestly: the 96% dormancy baseline needs internal validation before treating as ground truth
- •
RICE discipline catches feature bloat: both Job Fit Score and Application Readiness nudge had to earn their place through scoring, not assumption
What Surprised Me
How much of LinkedIn's early-career problem traces to structural incentives (experience-weighted matching, Premium paywalls) rather than missing features. The fix isn't adding more, it's reweighting what already exists.
What I Would Change
Run 8-10 structured interviews with early-career job seekers before writing the PRD. The personas are synthesized from secondary research, which is directionally useful but not a substitute for hearing real users describe the problem in their own words.
What's the actual baseline ECAR from LinkedIn's internal data? The 96% dormancy figure suggests low single digits, but that's a back-calculation, not a measured number. Also: does explaining a match help or hurt before the underlying match quality improves? That assumption sits underneath the decision to defer Match Explanation to Release 3.
- ?
How would users react to seeing only 15 jobs when they're used to hundreds? Would it feel limiting or liberating?
- ?
What if the algorithm gets it wrong? How do you build trust when you're filtering out 95% of options?
- ?
Could this work for senior roles where there are naturally fewer openings anyway?
PM Reflection
Personal thoughts on the product decisions and process
This case study demonstrates the full PM process from problem framing through business impact, built entirely from public secondary research. The biggest limitation, flagged throughout, is the lack of primary user interviews - the personas synthesize research patterns rather than quoting real people. What this version proves is the reasoning process: how a problem gets scoped, prioritized, and sequenced into a shippable release. The trade-offs section names five choices that closed off real alternatives, argued honestly in both directions. The hardest one: choosing to fix feedback (silence) before fixing matching (relevance), even though better matches feel more exciting, because closed-loop status has higher confidence and generates the data better matching would need. Every decision traces back to one core bet: LinkedIn is the only platform with the graph, reach, and data to solve early-career problems completely - it just hasn't pointed that machinery at this segment on purpose yet.