The Most Valuable Marketplace in Professional Life Is Waiting to Be Built

This week I was reading Summation and a piece by Auren Hoffman on why all recommendations suck. And, sadly, for the most part they do. The post also explicitly calls out LinkedIn's job recommendations. A prominent tech CEO gets suggested three roles with nothing in common. An accomplished lawyer sees a $16/hr suit attendant job and a $663K head of legal position in the same feed. The consensus: recommendations are broken, and nobody at LinkedIn is trying. That may be true, but this framing also potentially missed the real problem (and opportunity) LinkedIn has.

A recent conversation with another founder got me thinking about this more concretely. Now seems like the right time to formalize those thoughts. LinkedIn's job recommendations aren't just bad. They're symptomatic of a structural marketplace failure that's degrading value for everyone in the system.

The Problem Isn't the Algorithm. It's the Data Model.

LinkedIn has over a billion profiles, access to state-of-the-art language models through Microsoft, and thousands of engineers. So why are the recommendations still this bad?

Because no model — no matter how sophisticated — can generate meaningful job matches from the data LinkedIn actually collects. The core profile is built on three signals: where you worked, what your title was, and for how long. For almost all profiles, that's it. That's the input to a system that's supposed to answer one of the highest-stakes questions in a person's life: what should I do next?

Title and tenure tell you almost nothing about what someone actually did, what they're good at, or where they'd thrive. A "Senior Product Manager" at a 15-person startup and a "Senior Product Manager" at Google had fundamentally different roles, scopes, and skill sets. On the surface, LinkedIn treats them as equivalent data points.

This isn't a model problem. It's a data problem. And it has cascading consequences.

The Vicious Cycle

When recommendations are poor, job seekers can't find relevant roles through the platform. So they do what's rational: they apply broadly. Volume becomes the strategy because precision isn't available.

This floods employers with untargeted applications. Signal-to-noise collapses. Hiring managers and recruiters stop trusting LinkedIn as a sourcing channel, response rates drop, and the best roles increasingly get filled through referrals, warm introductions, and networks that exist outside the platform entirely.

Job seekers notice. They get fewer responses, so they apply to even more roles. The cycle accelerates. And it's not just a LinkedIn problem. It's become the defining feature of the modern hiring market. The primary path to employment for most professionals isn't a cold application through any platform. It's a warm connection. That's a damning indictment of every job marketplace, but especially the one with a billion profiles and the word "network" in its identity.

The result is a marketplace where neither side trusts the matching mechanism. Employers pay for access but rely on their own filtering. Job seekers use the platform because they're supposed to, but don't expect it to actually work. LinkedIn retains its position through network effects and incumbency, not because it's delivering real matching value.

This is a classic marketplace failure, and it's also a massive opportunity.

The Fix: Make the Profile Worth Building

The path forward requires LinkedIn to fundamentally rethink what a profile is. Not a static resume. Not a public-facing highlight reel. A living, continuously enriched record of what someone has actually done.

Today, LinkedIn incentivizes profile creation for multi-player value: you build your profile so recruiters can find you and connections can see your career and who you know. But that only motivates people to fill in the minimum — title, company, dates — because that's all the format rewards.

The unlock is in creating greater single-player value. Give people a reason to enrich their profile for themselves (that goes beyond keyword stuffing), even when they're not actively job searching.

Here's what this could look like in practice:

LinkedIn helps employed professionals continuously capture their work — projects owned, outcomes delivered, skills applied, impact measured — in a structured, private layer beneath the public profile. Think of it as a "Career Journal" that LinkedIn helps you maintain.

There are natural entry points for this. Prompt users on a recurring cadence (ideally timed to their employer's review cycle, or quarterly by default): What did you ship this quarter? What was the outcome? What tools or skills did you use? The user doesn't need to write a detailed account. A 90-second voice memo could be enough for AI to structure into meaningful career data. LinkedIn could ingest documents users already have — performance reviews, shipped work, role descriptions — pulled from email, cloud storage, or local files with minimal effort. It could even prompt contextually: when a user posts about a company win or engages with a team milestone, LinkedIn asks if they were involved and what their role was. Not to display publicly, but to build LinkedIn's understanding of what you've actually done.

This doesn't need to be public. In fact, keeping it private by default is the right design choice. The data layer sits beneath the profile. LinkedIn mines it for matching, but the user controls what surfaces. This addresses the obvious privacy concern while preserving the data utility.

The single-player value is immediate and tangible. Most professionals are notoriously bad at tracking their own accomplishments. When it's time to update a resume, write a self-review, negotiate a promotion, or prepare for an interview, they're reconstructing months or years of work from memory. LinkedIn could solve that problem today. It becomes the system of record for your career — not just where you worked, but what you built.

And when that user is ready for their next opportunity, LinkedIn has the context to do something no other platform can: generate a tailored resume, craft a narrative around their career arc, and identify roles where their specific experience is a genuine fit. The profile goes from a static document you dread updating to a living asset that compounds in value over time.

Why This Didn't Work Before, And Why It Can Now

None of this is a conceptually new idea. LinkedIn has almost certainly explored versions of richer profiles before. The likely reason it didn't work comes down to a basic behavioral equation: for a user to change their behavior, the perceived value of the effort has to outweigh the burden of the effort itself.

The value side was always there. A richer career record would lead to better matching, stronger narratives, and smarter career moves. But that value is temporally lagged — potentially by years. You're asking someone to invest effort today for a payoff they won't experience until their next job search. That's a hard sell, and it's even harder when the effort required is exhaustive manual data entry.

What's changed is the other side of the equation. AI has fundamentally collapsed the burden. The user's input can now be a 90-second voice clip, a document upload, or a reaction to a contextual prompt. The structuring, categorization, and synthesis happen on the backend. The ask goes from "spend hours maintaining a detailed career log" to "spend five minutes a year so the next five years of your career are better."

Ideally, you move both levers. A richer data asset does create more value for the user — better recommendations, stronger resumes, clearer career narratives. But the real unlock is lightening the burden to the point where adoption becomes almost effortless. The equation between "how much work is this for me" and "how much benefit will I get" has been fundamentally rebalanced.

The Virtuous Cycle

With richer candidate data, the entire marketplace dynamic inverts.

LinkedIn can now match candidates to roles based not just on title and tenure, but on actual work product, demonstrated skills, and measurable outcomes. Matching confidence goes up. Recommendations become genuinely useful, maybe not perfect, but directionally right. LinkedIn can even surface a confidence score: strong match, partial match, stretch role, giving both sides a shared vocabulary for fit.

When recommendations improve, applications become more targeted. Employers see higher-quality candidate pools. Response rates increase. The channel starts delivering real value, and employers invest more in it: posting more roles, using more premium features, trusting LinkedIn as a genuine talent acquisition funnel rather than a firehose they need to filter themselves.

Job seekers get responses, which reinforces the behavior of applying through the platform. The flywheel spins.

This also transforms what LinkedIn can offer recruiters. A friend who runs a recruiting practice put it bluntly: LinkedIn's search and recommendations are garbage for recruiters too. Today, even LinkedIn's most advanced recruiter tools are constrained by what profiles actually contain: titles, skills tags, and keywords. A recruiter can filter for seniority and industry, but they can't search for someone who has actually done specific work.

With structured work-product data, recruiters could search for candidates who have actually done specific things: led a product through regulatory approval, scaled a system from one market to five, built and managed a cross-functional team of a certain size. That's a fundamentally more valuable search product than what exists today, and it justifies meaningfully higher pricing. When the signal is that strong, the connection itself carries real value. A far cry from the current model where InMails go unanswered because neither side trusts the match.

Why LinkedIn Should Act Now

Two reasons.

The disruption risk is real. The gap between what LinkedIn offers and what AI makes possible is widening every quarter. A new entrant that builds a profile-and-matching system around rich work data — not just resume metadata — would have a compelling wedge. LinkedIn's moat is its network, but networks can be unbundled when a focused product delivers dramatically better outcomes for a specific use case. Job matching is exactly that kind of use case.

The monetization upside is significant. LinkedIn's current talent solutions business is built on a thin data layer. Richer candidate understanding means higher match confidence, which means LinkedIn can charge more for connections where the signal is strong. The economics of a high-precision matching marketplace are fundamentally different from those of a glorified job board.

The Opportunity

Helping people navigate their career is one of the most meaningful problems a product can solve. LinkedIn has every structural advantage to do it: the network, the data foundation, the distribution, the AI capabilities.

What they're missing isn't technology. It's a product vision that treats the profile as something worth building for its own sake, and the job marketplace as something worth making actually work.

Building a talent marketplace where both sides trust the match creates a durable, defensible moat around professional data, and that starts with understanding what people have actually done, not just where they did it. Better recommendations aren't the objective. They're a byproduct of getting the fundamentals right.

LinkedIn is best positioned to own this. But right now, they're leaving the door open for someone else to walk through it.