How LinkedIn grows

Premium members can mark up to three jobs a month as their top choice and add a two-sentence elevator pitch on why they are the best candidate. The recruiter sees that the candidate marked the job as a top choice, with the pitch. Designed so it helps job seekers stand out without paywalling applying and also benefits recruiters (a scarce, credible signal).

When a member enters the Premium flow, ask whether they are looking for a job, growing a business, hiring, or something else; the answer decides which content and plans are shown. Behavioral signals fill in too (e.g. a member who just searched for a job gets the job-seeker experience). Experiences are also tailored by country and industry. Machine learning drives it, moving toward LLM-based personalization.

A perks section in the Premium benefits dashboard. Monthly rotating perks (e.g. Notion with AI free for six months, three months of Spotify, four months of Headspace) plus evergreen perks rotated less often, every 6-12 months (e.g. three months of Duolingo, 'Washington Journal', coaching).

Every spec names the target member and the job to be done. A team dedicated to the value of the subscription (building new features, driving adoption of existing ones) is roughly the same size as the team driving growth through pricing, discovery, packaging and promotions. Over ~2 years they added 20+ Premium benefits: job-match insights (which jobs you're likely to hear back from, how you rank), profile standout features (rotating cover photos, highlighted achievements and recommendations, profile-fill help), events and learning courses. They then explain to members how to use each benefit.

Everything launched to market (new features, layouts, value messaging) goes through an A/B test. Bottoms-up culture: each PM owns an area/domain with its own goals and roadmap and runs their own tests. An experimentation system measures each test and detects overlaps; tests that could contaminate each other's reads are time-sequenced/staggered. Tests ramp if they win and de-ramp if not. Guardrail: experiences the team would not be proud of or that disrupt members are not even tested; ties go to the member.

Annual Premium plans get a significant discount. In test: dual plans covering you plus a family member, friend, partner, spouse or colleague, with 20% off for taking two plans. Otherwise Premium is not discounted (no coupon codes).

Identify which benefits correlate with retention (e.g. members who hear back from recruiters retain better) and build more value props around them. Improve messaging that tells members which benefits they have and how to use them.

Prior-subscriber status is used to predict likelihood to resubscribe and chooses which experiences and messaging a returning member sees. Win-back promotions for 'boomerangs' are shown only at times, based on signals, sometimes decided by AI algorithms. Goal is that people who leave leave happy so they come back when their needs change (new job search, starting a business).

Free forever: building a profile, posting (equally distributed in feed), finding and applying to jobs; nothing related to finding or applying for a job is paywalled because recruiters (the jobs business) need a healthy free member base. Paid: features that drive business outcomes, e.g. a custom 'Visit my website'/'View my newsletter' link that shows under your name on posts, messages and search. Freemium tiers of the same feature: one cover photo is free, multiple rotating cover photos (portfolio, recommendations) are paid.

North star is long-term revenue. Every ramp/initiative is measured on one-year bookings, then one-year revenue is modeled; for experiments whose effect changes over time (e.g. pricing) revenue is modeled on a five-year horizon and the decision is made on the five-year view. Supporting metrics: members acquired, trial-to-paid and paid retention, and 'Premium sessions' (sessions where a Premium member uses a Premium feature). Revenue that would squeeze the funnel is deliberately foregone.

Because Premium changes a lot over time, even former paid subscribers can be offered another free trial. Anyone who had a free trial in the last 12 months pays; after 12 months they are testing different scenarios for when and how to re-offer the trial. (Host notes app stores block repeat trials on the same product unless using offers.)

LLM features solve specific jobs: writing a profile summary (three generated options from profile data, editable, with shorter/longer/more or less professional controls), drafting job descriptions for small-business hirers, ranking candidates, short comment starters. They are positioned as help with the member's task rather than as 'AI' everywhere.