Skip to main content

What are you looking for?

Explore our services and discover how we can help you achieve your goals

Marketplace platform architecture: vendors, catalog, payments, AI and scale

A marketplace platform adds four things a single-seller store lacks: vendor onboarding and accounts, a catalog that merges many sellers' listings, a money layer that splits payments and pays vendors out, and order routing per vendor. Build them as modules around a standard commerce core, use a marketplace payments provider rather than moving money yourself, and add AI search, ranking and fraud signals once listings and orders supply the data.

Book a solution review See the related service

Reviewed by David (CEO) · Updated 17 Sep 2026 · 11 min read

star

This technical article is for CTOs, architects and product owners planning a multi-vendor marketplace or turning a store into one. It explains the architecture, the main trade-offs and how to test it. It is informational: if you are ready to scope a build, the e-commerce marketplace solution is the place to start. For the wider question of when a custom platform is justified at all, read our custom e-commerce platform guide.

Context: what changes when a store becomes a marketplace

A single-seller store has one catalog owner, one merchant and one fulfilment operation. A marketplace has many of each, and the platform sits between them.

Concern Single-seller store Marketplace
Who sells The store owner Many independent vendors
Catalog One team maintains it Vendors supply listings; the platform curates them
Money Payment goes to the merchant Payment is split between vendors and the platform, then paid out
Orders One order, one fulfilment One basket can become several vendor orders
Trust The brand vouches for itself The platform must verify vendors and handle disputes
Compliance The merchant's obligations Added platform duties, such as reporting seller income in some regions

Every row adds data, rules and failure modes. The architecture should make each of them explicit rather than bending a single-seller store until it breaks.

Reference architecture

  • Storefront

    Responsibility
    Browsing, search, basket and checkout for shoppers
    Design note
    Can be the commerce platform's own front end or a headless front end
  • Commerce core

    Responsibility
    Products, prices, baskets, orders, customers
    Design note
    Keep it standard; marketplace logic lives around it
  • Vendor portal

    Responsibility
    Vendor sign-up, profile, listings, orders, payouts, reports
    Design note
    Vendors act only on their own records
  • Onboarding and verification

    Responsibility
    Collects vendor identity, business and bank details
    Design note
    Delegate identity and payout verification to the payment provider
  • Catalog and listing pipeline

    Responsibility
    Imports, validates, moderates and publishes listings
    Design note
    Queue-based, so a large vendor feed does not slow the store
  • Search

    Responsibility
    Indexes listings across vendors
    Design note
    A dedicated search engine, fed from the catalog, with room for semantic search and learned ranking
  • Payments and ledger

    Responsibility
    Charges shoppers, splits amounts, records commissions and fees
    Design note
    The ledger is the source of truth for every split
  • Payouts

    Responsibility
    Moves vendor balances to vendor bank accounts
    Design note
    Scheduled, with holds for refunds and disputes
  • Order routing

    Responsibility
    Splits baskets into vendor orders and tracks each one
    Design note
    Status rolls up to one view for the shopper
  • Back office

    Responsibility
    Moderation, disputes, commissions, reporting
    Design note
    Role-based access for platform staff

An order's path through this architecture:

  1. The shopper checks out a basket containing items from three vendors.
  2. The payment provider charges the shopper once.
  3. The ledger records each vendor's share, the platform commission and any fees.
  4. Order routing creates three vendor orders and notifies each vendor.
  5. Each vendor ships and updates its own order; the shopper sees one combined status.
  6. After the holding period, payouts move each vendor's balance to its bank account.
  7. A refund or dispute on one vendor order adjusts only that vendor's balance.

Vendor onboarding and accounts

Onboarding speed decides how fast the catalog can grow, so treat it as a product flow, not an admin form. A good flow collects only what is needed to start, lets the vendor list products while verification runs, and blocks payouts, not listings, until verification is complete.

Two parts should not be built from scratch:

  • Identity and payout verification. Marketplace payment services handle this. Stripe Connect, for example, provides onboarding options for connected accounts, collects the verification information each account needs, and manages payouts to them.
  • Seller reporting duties. In the EU, DAC7 (Council Directive (EU) 2021/514), which has applied since 1 January 2023, requires digital platform operators to collect and verify information about their sellers and report it annually to tax authorities. Design the vendor data model so these fields are captured at onboarding, not retrofitted.

Vendor accounts need their own roles: an owner, staff who manage listings, and staff who see orders and payouts. Every vendor query must be scoped to that vendor's records.

Catalog: vendor listings or a shared product catalog

Model How it works Strengths Trade-offs
Vendor-owned listings Each vendor creates its own product pages Simple; fast onboarding Duplicate products, uneven quality, weaker search
Shared catalog with offers One product page, many vendor offers with their own price and stock Clean search and comparison Needs product matching and catalog governance
Curated listings Vendors submit, the platform edits and approves Consistent brand experience Moderation capacity limits growth

Most fashion and lifestyle marketplaces lean towards vendor-owned or curated listings; marketplaces for identical goods lean towards a shared catalog with offers. Whichever model you choose, run listings through a pipeline: import, validate required attributes and images, moderate, then publish to search. Queues keep a large import from affecting shoppers.

Payments and payouts

The money layer is where marketplaces carry the most risk. Three decisions shape it.

Decision Options What to weigh
Who is the merchant of record The platform, or each vendor Liability for refunds, disputes and tax; how the provider structures charges
How the charge is split At charge time, or by transfers after the charge Flexibility for multi-vendor baskets versus simplicity
When vendors are paid On a fixed schedule, or after delivery confirmation Cash flow for vendors versus exposure to refunds and disputes

Use a payment service designed for platforms rather than collecting money into your own account and paying vendors manually. Stripe Connect is one example: it lets a marketplace collect payments from customers and pay a portion out to sellers, with connected-account balances and payouts managed by the provider. Keep your own ledger anyway, because commissions, fees, refunds and adjustments must reconcile against the provider's records.

Card data should stay out of your systems. PCI DSS, the Payment Card Industry Data Security Standard, sets baseline technical and operational requirements for any entity that stores, processes or transmits cardholder data. Using the provider's hosted payment fields or checkout keeps most card handling on the provider's side and reduces your own compliance scope.

Orders, fulfilment and returns

  • Split early. Create vendor orders at checkout, not later, so each vendor sees its work immediately.
  • Shipping per vendor. Each vendor may ship from a different place with different rates and times; show the shopper one delivery summary built from them.
  • Aggregate status. The shopper sees one order with per-vendor lines; support staff see the full trail.
  • Returns and refunds per vendor line. A return on one vendor's item must not touch another vendor's balance.

Scaling the platform

  • Catalog growth

    Symptom
    Slow search and category pages
    Architecture response
    A dedicated search index fed from the catalog pipeline
  • Large vendor imports

    Symptom
    Store slows during imports
    Architecture response
    Queued, batched imports with back-pressure
  • Checkout peaks

    Symptom
    Timeouts at checkout
    Architecture response
    Stateless application servers, caching for catalog reads, idempotent payment calls
  • Vendor count

    Symptom
    Onboarding queue grows
    Architecture response
    Self-service onboarding with delegated verification
  • Reporting load

    Symptom
    Back-office reports slow the store
    Architecture response
    Separate read models or a reporting database
  • Order volume

    Symptom
    Delayed notifications and payouts
    Architecture response
    Event-driven processing for routing, notifications and ledger updates

Choosing the platform base

Netbase has delivered commerce platforms on WooCommerce, Magento 2, Laravel and headless commerce, and its wider stack includes PrestaShop, OpenCart, Shopware, CS-Cart, Shopify, Salesforce, Akeneo, Odoo and Symfony. For a marketplace, the choice usually comes down to three options:

Base When it fits Trade-offs
Commerce platform with marketplace extensions (for example WooCommerce or Magento 2) Standard marketplace flows, a need to launch quickly Extension limits; custom rules can fight the platform
Custom application (for example Laravel) Distinctive vendor, catalog or payout rules More to build and maintain; full control
Headless commerce with separate marketplace services Several front ends, high traffic, a need for independent scaling Most moving parts; needs strong engineering operations

Whichever base you choose, design it API-first, so vendor apps, partner feeds and AI services use the same interfaces as the storefront.

Security and trust

Marketplaces multiply the number of people with access. Apply role-based access control per vendor and per platform role, MFA for admin and vendor dashboards, TLS in transit and encryption at rest, and regular vulnerability scanning. Netbase delivery follows these practices, together with secure code review, penetration testing and disaster recovery planning. Add marketplace-specific controls: fraud checks on new vendors and unusual payout changes, moderation for listings and reviews, and an audit trail for every change to bank details. AI risk scores can order the review queue, but a person decides on suspensions and payout holds.

Test evidence: what to test before launch

These are general engineering practices, not results from a specific project:

  • Ledger reconciliation tests. For generated baskets across several vendors, commissions and fees, confirm that shares, refunds and payouts balance to zero against the provider's records.
  • Split-order tests. Baskets with one, several and many vendors, including partial refunds and cancellations on one line.
  • Access tests. Automated attempts by one vendor to read or change another vendor's listings, orders or payouts must fail.
  • Import load tests. Run the largest expected vendor feed during simulated shopper traffic and check that page response times hold.
  • Payout failure tests. Rejected bank details, payout holds and disputes must leave balances correct and visible.
  • Ranking and risk checks. AI search and fraud scores are compared against a labelled sample of queries and past cases before they influence what shoppers see or which vendors are held.

AI in a marketplace platform

AI is an addition to a working marketplace, not a launch requirement: listings, searches and orders from many vendors give it the data to learn from. Three uses matter most. AI search and ranking match shopper intent across inconsistent vendor listings and balance relevance with vendor fairness. Listing enrichment and moderation draft missing attributes, match duplicates to a shared product and flag prohibited items for a moderator. Fraud signals score new vendors, sudden bank-detail changes, fake reviews and unusual refund patterns, so staff review the riskiest cases first. Each item below states how mature it is at Netbase.

Example: an EU fashion-tech marketplace

Netbase built a multi-vendor marketplace for an EU fashion-tech aggregator moving from direct sales to a marketplace model; the client stays anonymous. The published case reports gross merchandise value (GMV) up 47% and vendor onboarding time down 60%, both within six months of launch. It supports two points from this article: onboarding is a growth lever, and a marketplace payments provider (Stripe Connect in this case) lets engineering effort go to onboarding and catalog flows.

Read the marketplace case

Two more records show the same parts in other markets. RB Marketplace, for West Africa and the diaspora, pairs vendor approval, commissions and payouts with a customer app on the marketplace API. An online classifieds platform, for a founder who is not named, adds AI filtering that flags offensive ads for admin moderation. Neither record publishes a result.

Plan the next step with a Netbase consultant

Limits of this article

  • It describes a reference architecture and general practice; every marketplace adjusts it to its vendors, goods and regions.
  • Payment, tax and consumer rules differ by country. DAC7 and PCI DSS are cited as examples of obligations to design for, not as a complete list; take legal and tax advice.
  • Stripe Connect is named as one example of a platform payments service, not as a recommendation over other providers.
  • The EU marketplace figures are that client's reported results for six months after launch; absolute values are not published.

Next step

Share your vendor model, catalog size and payment regions, and we will book a solution review to sketch the architecture and the first release. You can also see the related service, see how Netbase works in retail and e-commerce, or browse more Netbase insights.

AI-enabled ecommerce development for merchants who outgrew templates AI-enabled ecommerce development for merchants who outgrew templates

Netbase provides custom ecommerce development for merchants that have outgrown templates, to turn existing traffic into orders and run operations with less manual work, using AI for search, recommendations and catalog work where it pays. We build on WooCommerce, Magento 2, Laravel and headless stacks; 4over4, whose store gained an AI recommendation engine, reported 82% more revenue within six months.

Learn More
line
Multi-vendor marketplace development: vendors, catalog, payouts and AI search on one platform Multi-vendor marketplace development: vendors, catalog, payouts and AI search on one platform

A multi-vendor marketplace is a commerce platform where many independent sellers list, sell and get paid through one storefront. Netbase's marketplace solution covers vendor onboarding, a shared catalog, split payments and payouts, with AI for search, listing enrichment and fraud checks. For an EU fashion-tech marketplace, Netbase's work grew gross merchandise value (GMV) 47% and cut vendor onboarding time 60%.

Learn More
line
Contact Netbase

Discuss a project

Netbase JSC helps organizations design, build, modernize, and operate digital products and AI-enabled business systems.
Project enquiries

[email protected]

WhatsApp

+84 937 869 689

Office address

91 Nguyen Chi Thanh, Dong Da, Hanoi, Vietnam

Get in touch

Tell us what you want to build, modernize, or operate.

Tell us what you want to build, modernize, or operate.

Contact Netbase