Unbundling the Marketplace Stack

The marketplace model is structurally undermined (see essay Thoughts on the Internet Marketplace for the high level analysis). There will be an 'unbundling' of all the specific business functions the marketplaces currently solve for.

The table below summarizes the current and future stack across the various layers. For two decades, the marketplace had pricing power because they owned every layer. However, as more efficient ways for discovery emerge, there will be opportunities for new companies to fill in the gaps for a fraction of the cost.

Further analysis on each layer below. As discussed, the catalyst is discovery, and the other layers follow, so there are gaps (or opportunities -- especially in trust, escrow, and relationship) in this new paradigm.

Customer acquisition: Marketplaces position customer acquisition as a service they provide to sellers. The reality is more extractive. Most established marketplaces spend 15–25% of gross bookings on sales and marketing. The majority of this goes toward demand generation for the marketplace brand, rather than any individual seller. The KPI for any marketing campaign Fiverr runs, for example, is brand awareness and bookings growth. Sellers surrender a meaningful percentage of revenue so the platform can run campaigns that build its own brand equity. As direct discovery becomes viable through frontier models, social media, and owned channels, sellers will shift toward tools that deliver measurable outcomes for their own business. Customer acquisition becomes a direct relationship: the seller pays for results, keeps the customer data, and owns the repeat booking.

Discovery: Frontier models and social media platforms will be the demand aggregators. The unit economics of aggregating demand for all purposes (rather than a specific vertical) is just far better. Additionally, these platforms know the buyer at a level that marketplaces don't. Users will increasingly connect their email, calendars, and more -- giving these platforms an extreme personalization advantage by default.

Profile: Today the seller's profile and data related to it (reviews, booking history, etc) belongs to the platform. Your Airbnb Superhost status does not transfer to Vrbo. Sellers rebuild trust from scratch on every new platform. This helps marketplaces maintain leverage, but puts sellers in a position of weakness. There will be an agentic business profile -- whether its on different platforms or as an abstraction layer -- that allows sellers to truly own their online presence.

Trust: The most difficult layer to replace because a huge part of it is simply brand reputation. This reputation signals that a seller delivers what they promise. Today it lives in platform reviews, walled off deliberately to protect the marketplace's data moat. This is already breaking: frontier models synthesize richer trust signals than any single platform's review system, pulling from social graphs, open review sources, and verified transaction history. Reputation will evolve into something resembling a credit score — portable, longitudinal, continuously updated, and owned by the seller. The marketplace that walls off its reviews will find the moat drying up as buyers stop looking there first. Guarantee is different, and is addressed in escrow below.

Scheduling: Scheduling was never a marketplace's core competency. It was table stakes — something the platform had to provide to make the bundle function, and handle rescheduling and cancellations. Once the seller's profile is portable and discovery happens outside the platform, there is no reason for scheduling to route through a marketplace. Dedicated calendar APIs already handle availability, booking logic, and conflict resolution more cleanly than any marketplace-native tool. Sellers will connect their calendar directly, and any compatible AI agent will book against it without the marketplace ever touching the transaction.

Payments and compliance: Payment processors already run global compliance, fraud detection, chargebacks, and tax at scales no vertical marketplace can match. Stripe's Agentic Commerce Protocol, as well as other protocols, make the connection explicit: a seller links their payment credentials once, and any compatible AI agent can process the transaction directly. From payments, these processors expand naturally up the value chain into accounting automation, tax compliance, and customer management. The seller's financial infrastructure becomes richer and more capable than anything a marketplace offered, at a fraction of the cost. The intermediary is removed from the flow, and with it, the justification for the cut it was taking.

Escrow: Some buyers will always want protection against a bad outcome — the listing isn't as described, the freelancer disappears, the service isn't delivered. Today that guarantee is bundled into the take rate whether the buyer needs it or not. The host with a perfect ten-year track record pays the same blended rate as a new listing with no history. The honest version of this service is optional insurance: a small fee — 2–4% of the transaction — chosen by both parties. The seller opts in to offer it as a quality signal. The buyer opts in to purchase it as downside protection. Neither pays for risk they are not taking. The company providing the guarantee has one job: resolve disputes fairly, and its entire business model depends on doing that well. This is the clearest greenfield in the new stack — a standalone service guarantee layer, not bundled into a platform take rate, priced proportionally to actual risk.

Relationship: The customer relationship is the compounding asset that marketplaces have quietly captured. Every booking that routes through a platform adds to the platform's understanding of the buyer and zero to the seller's. The seller learns nothing — no contact information, no booking history, no ability to reach back out. When the buyer returns, they return to the marketplace, not to the seller. As direct channels become viable, this dynamic inverts. The seller owns the communication, the rebooking, the referral. The relationship compounds on the seller's side of the ledger instead of the platform's. This is not a technical shift — the tools already exist. It is a behavioral one, and it follows naturally once the discovery and payment layers no longer require a marketplace intermediary.