Service Marketplaces Are Still Mostly Offline — And That's the Opportunity

A16Z partners Li Jin and Andrew Chen published a framework mapping the four historical eras of service marketplaces — from Craigslist-style listings to managed marketplaces like Opendoor — and argued that the next major wave will be marketplaces that unlock regulated service industries. They identified five distinct strategies for doing so, including hiring supply directly, leveraging geographic arbitrage, using unlicensed providers, and applying AI to automate licensed tasks. The core thesis: the $9.7 trillion U.S. consumer service economy is still mostly offline, and that gap represents the next generation of trillion-dollar marketplace companies.

Most marketplace founders benchmark themselves against goods marketplaces — Amazon, eBay, Etsy. But services are structurally different: they're produced and consumed at the same time, quality is subjective, and supply is often fragmented or legally constrained. This isn't a technology gap — it's a structural one. The reason services resisted digitization isn't that founders weren't trying; it's that each era of marketplace model was only suited to a narrow band of service complexity. The insight here is that marketplace model design must match the complexity of the service. A listings model works for simple, low-trust transactions. A managed model is required when stakes are higher. Regulated services require an entirely different set of supply-unlocking strategies. Founders who want to build a successful marketplace must avoid treating all service categories as equivalent, or they risk building the wrong platform for their market.

SUPPLY: In service marketplaces, supply is the hardest problem. Unlike goods, you can't warehouse it or manufacture more. Supply is human, local, fragmented, and often legally constrained by licensing. Each of the five strategies in the article is fundamentally a supply-side intervention — not a demand-side one. This tells you where leverage actually lives in service marketplaces. DEMAND: Demand in services is abundant but latent. People want better childcare, legal help, and home care — but they default to informal referral networks because platforms haven't earned enough trust to replace word-of-mouth. Unlocking demand is downstream of solving trust, which is downstream of solving supply quality. LIQUIDITY: Service marketplaces face a harder liquidity problem than goods marketplaces because transactions are time-bound, location-bound, and identity-dependent. You can't substitute one provider for another the way you can substitute one product SKU for another. This means liquidity requires either dense local supply or a model that decouples geography from delivery (remote services). TRUST: Trust is the core mechanic in every era described. Each new marketplace generation — from listings to managed models — is essentially a new trust architecture. The managed marketplace era shifted trust from the provider to the platform. This is the key design decision: who is accountable for quality, the platform or the individual provider? GROWTH: Service marketplaces that nail a single high-frequency use case first (rideshare, food delivery) can expand later. Those that try to serve complex, low-frequency services first face a much harder growth path because the feedback loop is slow and word-of-mouth takes longer to compound. ONBOARDING: Supply onboarding in regulated services is not a UX problem — it's an operational and compliance problem. Friction in onboarding licensed providers is often structural (credentialing, background checks, insurance requirements). Founders who treat this as a product problem will underinvest in the operational infrastructure needed to onboard supply at scale. MONETIZATION: Managed marketplaces justify higher take rates because they absorb more operational cost and deliver more guaranteed outcomes. The tradeoff is margin compression. Founders choosing between a light-touch marketplace and a managed model are implicitly choosing between scale-at-low-margin and quality-at-high-margin. Neither is wrong — but the choice must be made deliberately based on what the service category actually requires, alongside a clear understanding of marketplace launch best practices from the outset.

• Diagnose your supply constraint before anything else. Is supply scarce because providers don't know about you, don't trust you, aren't licensed, or are geographically mismatched with demand? Each problem requires a different intervention.

• Match your marketplace model to your service's trust requirements. High-stakes, high-price, subjective services (healthcare, legal, childcare) require managed models. Commodity, binary-outcome services (delivery, transportation) can run on lighter-touch models.

• If you're in a regulated category, map the licensing landscape before building. Understand which states require which licenses, how long certification takes, and what the drop-off rate is in your supplier funnel due to compliance friction.

• Consider geographic arbitrage as a supply strategy. If your service can be delivered remotely (design, tutoring, legal advice, financial planning), your addressable supply pool is national or global — not local. This changes your liquidity math entirely.

• Don't treat trust as a feature — treat it as the product. Every element of your marketplace (reviews, vetting, guarantees, insurance, branding) is a trust mechanism. Map out who is accountable for quality at each stage of the transaction and design accordingly.

• If you're considering a managed model, pressure-test your unit economics early. Managed models create better customer experiences but compress margins. Know your cost per transaction before you scale.

• Use the 'who owns the customer relationship?' question as a design test. If customers think of your marketplace as the service provider (like they do with Uber), you have more control over quality but more operational liability. If they think of the individual provider, you have less control but lower operational burden. Choose deliberately.