Credentialing is a two-sided market — and nobody builds the supply side

Credentialing is treated as a verifier problem. How do employers screen? How do candidates signal? Which institution’s name carries enough weight to short-circuit the actual evaluation of skill? The entire discourse — from signaling theory to ATS parsing to “is this degree from a good school” — orbits one side of the market. And the entire industry, from HR-tech to the university brand economy, serves that same side. The verifier side is saturated. The supply side — the machinery that produces credible, portable, revocable, evidence-backed proof of skill — is barely built. Nobody has assembled it as a coherent platform. That gap is where the leverage sits.

The frame: credentialing is two-sided, and only one side has infrastructure

Credentialing is a two-sided market. The two sides are not “learners and employers,” which is the lazy framing. The two sides are issuers and learners on the supply side — the parties who produce credible evidence of skill — and verifiers on the demand side — the parties who consume that evidence to make decisions. Employers are the most visible verifier, but so are admissions offices, licensing boards, insurers, contracting platforms, and increasingly autonomous agents making routing decisions about human capability.

Classic two-sided markets tip to whoever solves the harder, underserved side. Payment networks solved merchant acquisition before they optimized the consumer card. Ride-sharing solved driver supply; the rider experience was the easy part. Marketplaces tip when the supply side reaches density, not when the demand-side UX gets slick. The side that’s hard to build is the side that becomes the moat, because competitors can copy your demand-side interface in a sprint, but they cannot copy your supply-side infrastructure without years of operational compounding.

Credentialing has never been framed this way. The coverage treats it as a signaling problem — a verifier-side approximation problem — and the capital flows accordingly. Nobody maps it as a two-sided market with the supply side missing. That mapping is the first move, because once you see it, the opportunity becomes structural rather than incremental: the supply side is unbuilt, and whoever builds it as a coherent platform owns the aggregation point the entire market routes through.

The verifier side is over-served

The demand side has had decades of capital and tooling. The modern hiring stack — ATS, resume parsing, skills ontologies, structured interview platforms, assessment providers selling to employers, background checks, reference automation — is a sprawling industry devoted to one thing: helping verifiers make cheaper decisions about candidates they cannot directly evaluate. Signaling theory, the dominant academic frame for credentials, is itself a verifier-side heuristic: it says credentials work because they communicate something cheaply to a party that lacks the time or expertise to assess directly. The university brand is a verifier-side proxy. “Is this a good school?” is a question a verifier asks to avoid the harder question: “Can this person do the work?”

This side is saturated. There are more tools for parsing, ranking, filtering, and screening candidates than any verifier can adopt. The buyer — the employer, the HR department — has been served continuously since the 1990s. Every cycle produces a new layer of verifier-side optimization: keyword matching, then semantic matching, then skills-graph inference, now LLM-assisted screening. None of it touches the underlying problem, which is that the evidence the verifier is trying to consume is thin, stale, and untrustworthy. You cannot build better verification on top of credentials that were never produced with verification in mind. The verifier side is polishing a lens while the object is out of focus.

The over-served verifier side is also why credentialing innovation feels stuck. The discourse cycles through the same debates — signaling vs. human capital, degree inflation, skills-based hiring, the “is college worth it” loop — because all of them are verifier-side questions. They ask what the consumer of a credential should do with it. They never ask how the credential was produced, whether the evidence inside it holds, or whether it will be valid in five years. The supply side is not in the conversation.

The supply side is unbuilt

The machinery that produces a credible, portable, revocable, evidence-backed credential is barely built. It exists in fragments — a proctoring vendor here, a badge format there, a wallet spec somewhere else — but nobody has assembled these into a coherent supply-side platform with a spine. The components themselves are underdeveloped:

Assessment that holds under generative AI. The entire assessment layer was designed for a world where producing a plausible answer required human effort. It no longer does. An assessment that cannot distinguish between a human who holds a skill and a model that can simulate its output is not a credential’s foundation — it is a liability. Assessment validity is the security perimeter of the supply side. Without it, the evidence produced downstream is forged by default. (Assessment validity is the new security perimeter.)

Issuer key custody. A signed credential is only as trustworthy as the key that signs it. If the issuer’s signing key is held in a shared spreadsheet, a SaaS vendor’s database, or a custody model where the issuer does not actually control the act of signing, the signature is theater. The custodian problem — who holds the key, under what SLA, with what separation between the issuer’s authority and the platform’s operation — is the trust foundation of the supply side. Most badge and wallet implementations hand-wave it. (The custodian problem.)

Revocation with a freshness SLA. A credential that cannot be revoked is a credential that cannot be trusted to mean anything at the time of verification. Most credentialing systems treat revocation as an afterthought — a revocation list that may or may not be checked, with no freshness guarantee, no verifier-side enforcement, and no issuer-side discipline. A supply-side platform has to treat revocation as a first-class capability with a defined SLA: how stale can a verification be before it must be re-resolved?

Portability across wallets and registries. A credential locked to a single platform is not portable; it is a hostage. Learners should not lose evidence of their skill because they switched employers or the issuer changed LMS vendors. Portability is a supply-side property — it is produced by the issuer’s choice of format, custody model, and publication surface. The verifier side cannot fix unportable credentials; it can only work around them, usually by ignoring them.

Agent-readable evidence. The next wave of verifiers is not human. Autonomous agents — hiring copilots, procurement agents, licensing bots — will resolve credentials programmatically. If the evidence inside a credential is not machine-readable in a way that preserves its cryptographic provenance and its semantic structure, it will be invisible to the verifier layer that matters most in five years. The discovery layer — how an agent finds and resolves a credential without browsing a catalog — is where the supply side meets the demand side at machine speed. (The discovery layer breaks.)

Nobody has assembled these five as a coherent platform. The market has fragments: a proctoring company that does not think about custody, a wallet vendor that does not think about assessment validity, a badge standard that does not think about agent readability, an ATS that does not think about the supply side at all. There is no spine. The supply side is not a market yet — it is a collection of disconnected components. Each component is a reasonable business in isolation — a point solution serving a sliver of the pipeline. But none of them assemble into a spine. No one ships the whole thing because no one is building the whole thing; everyone is building the piece they can sell into an existing procurement cycle. The result is a landscape of fragments, not a platform. And the market knows it — which is why every credentialing conversation eventually stalls at “we have standards, we just don’t have adoption.” The standards aren’t the problem. The missing platform is.

Why nobody builds it

The supply side is harder, less visible, and has no consumer glamour. VCs fund verifier-side UX — hiring tools, sourcing platforms, talent-intelligence dashboards — because that’s where the buyer sits with a budget and a quarterly hiring plan. The verifier side is a B2B SaaS motion with a clear ICP and a short sales cycle. The supply side is infrastructure. Plumbing. It’s assessment engines and custody layers and revocation registries and evidence schemas — the kind of stack that makes a partner’s eyes glaze over in a pitch deck and makes a CTO’s eyes light up three years too late. Infrastructure is unsexy and capital-thin until a platform owns it, and then it becomes the most valuable thing in the market and everyone wonders why nobody built it sooner.

This is the structural pattern I’ve written about before: the supply side is a gap market that institutional capital dismisses by default. The dismissal isn’t incidental — it’s the mechanism. The thing nobody wants to build is the thing nobody can quickly copy. The unsexy, capital-thin, operationally heavy middle of the stack is precisely where durable moats form, because the moat isn’t a feature — it’s the accumulated integration work that every fragment-player has structurally opted out of. When the gap is wide enough that capital doesn’t pursue it, the gap itself becomes the defensibility.

Own supply, own the platform

In a two-sided market, the underbuilt side is the leverage point. Whoever builds credible evidence-production as a coherent platform — assessment that holds under generative AI, issuer key custody that survives institutional churn, revocation with a freshness SLA, portability across wallets and registries, agent-readable evidence that a verifier’s automated pipeline can actually resolve — becomes the aggregation point that verifiers route through. Not because verifiers want a new vendor, but because verifiers want signal they can trust at machine speed, and the only way to get that is through supply-side infrastructure that produces it coherently.

This is the credentialing-specific instance of aggregation theory for credentials: aggregation is owning the interface to the underserved side. The verifier side can’t be the moat because everyone already serves it — ATS vendors, HR-tech platforms, signaling startups, brand-proxy providers, all competing on the same margin-compressed demand-side surface. The supply side can be the moat because nobody does. The asymmetry is the opportunity. You don’t win credentialing by building a better hiring tool. You win it by building the substrate that makes credentials worth resolving in the first place.

The honest cold-start

This is where the argument has to get honest, because the supply-side platform thesis is not a clean linear bet. Building supply is operationally heavy and slow. A two-sided market only tips when both sides show up — and verifiers won’t route through supply-side infrastructure until the credentials flowing through it are dense enough to be worth resolving. Until that density exists, you’re running infrastructure for credentials that verifiers don’t yet look for, issued by issuers who don’t yet see the point, into wallets that don’t yet have a reason to hold them. There is a real cold-start. It is not a feature gap you can ship your way through in a sprint. It is a liquidity problem — credential liquidity, verifier liquidity, issuer liquidity — and liquidity problems resolve through compounding, not through launches.

The bet is that the supply-side platform compounds once a critical mass of evidence-backed credentials exists. Issuers issue because the infrastructure makes issuance cheaper and more credible than the alternative. Verifiers resolve because the credentials are dense enough to justify the routing. Learners carry them because both sides recognize them. The flywheel is real — but it doesn’t spin from nothing. Until that mass accumulates, the work is heavy lifting with delayed payoff. You build assessment infrastructure that nobody queries yet. You run custody operations for issuers who are still testing the model. You maintain revocation freshness for credentials that no verifier has automated a check against. It’s not easy, and anyone who sells it as easy is selling you a fragment, not a platform.

Coda

The credentialing conversation keeps cycling — signaling, degree inflation, skills-based hiring, the death of the transcript, the resurrection of the transcript — because it keeps staring at the over-served side. Every debate is a verifier-side debate. Every startup is a verifier-side startup. Every framework is a verifier-side framework. The leverage is the supply side. Build it, and you own the platform the whole market routes through. Don’t build it, and you keep polishing a lens while the object stays out of focus.