Building a Marketplace or B2B Store: Architecture, Costs and What We Learned

To build a multi-vendor marketplace or B2B store, first check whether Shopify already covers you; if not, design seller isolation, stock, pricing and payment rules on the server before the storefront. We learned this building TheSkinProof, the founder's own marketplace, built and run by RAITHub, and Sundor Skin, a gated B2B wholesale platform built for a skincare importer.

This is a guide, not a sales page: the decisions, the hard problems, the build order, timelines and cost drivers, with every number named against the platform it comes from. If you are looking for a team to build yours, the service page is custom ecommerce and marketplace development.

What kinds of ecommerce platform need a custom build?

The ones where the rules are the product: many sellers in one store, verification workflows, trade pricing, credit, and fulfilment rules that an off-the-shelf platform cannot model well. Two kinds come up most.

• B2C multi-vendor marketplaces. Many sellers, one storefront, one checkout. Our example is TheSkinProof: 5 role-based portals (store, admin, seller, warehouse, auth), 217 API endpoints and 225K+ lines of code.
• B2B ordering and wholesale platforms. Approved buyers only, each with their own price and credit terms. Our example is Sundor Skin: 77 route pages, 146 PostgreSQL tables, 12 staff roles and 88 permission codes.

Both are live: TheSkinProof at theskinproof.com since Q2 2025, still in active development, and Sundor Skin at wholesale.sundorskin.com since September 2026. TheSkinProof is the founder's own venture rather than a client project, so read it as evidence of engineering, not of a client relationship. Full detail is in the TheSkinProof case study and the Sundor Skin case study.

When is Shopify the better choice than a custom build?

For most single-brand stores selling standard products at standard prices, Shopify or a similar hosted platform is the better choice. A custom build pays for itself only when your business rules are the product.

A hosted platform gives you checkout, hosting, themes, payment processing and an app ecosystem on day one, and someone else keeps it patched. If your catalogue, pricing and fulfilment look like most other stores, paying engineers to rebuild all of that is money you could spend on stock and marketing.

Custom starts to win when the rules that make your business different would otherwise live in plugins, spreadsheets and workarounds. This is the test we use.

Your situation	Usually the better fit	Why
One brand, a standard catalogue, card payments, a normal courier	Shopify or similar	Everything you need already exists and is maintained for you.
You are still testing whether people will buy	Shopify or similar	Validate demand before you pay for custom engineering.
Many independent sellers, each managing their own products, stock and orders	Custom, or a dedicated marketplace product	Seller isolation and seller tooling are the core of the business, not an add-on.
Every product must pass checks before it can be sold	Custom	Verification is a workflow with evidence and approvals, which store plugins rarely model well.
Trade buyers with tier prices, per-buyer overrides and credit limits	Custom, or a B2B-capable platform plan if your rules are simple	Credit and pricing rules have to be enforced on the server, not trusted to a theme.
Local payment rails, cash on delivery, or couriers that hosted plugins cover poorly	Custom	You need to own the integration and what happens when it fails.
You already run a store and it mostly works	Keep it	Replace only the part that is costing you money. A diagnostic will tell you which part that is.

The longer comparison is in Shopify vs custom ecommerce.

What are the hard problems in marketplace and B2B ecommerce?

The hard problems are the ones where a mistake costs money or trust. Seven come up again and again.

Hard problem	What goes wrong	How we solved it	Proof
Seller and buyer isolation	Someone reads or edits data that is not theirs	Separate role-based portals; PostgreSQL row-level security; permission codes checked on privileged actions	TheSkinProof: 5 role-based portals. Sundor Skin: row-level security, 12 staff roles, 88 permission codes
Product authenticity	Counterfeit or expired stock is listed	A four-gate verification pipeline: invoice, physical, ingredient, batch code	TheSkinProof
Cash-on-delivery fraud	Fake or refused orders waste stock and courier trips	Cash-on-delivery orders are risk-assessed before they are accepted	TheSkinProof
Stock correctness	Two customers buy the last unit, or older batches expire on the shelf	Per-variant stock decremented inside the order transaction; batch-level allocation, earliest expiry first (FEFO)	TheSkinProof (no overselling); Sundor Skin (FEFO)
Price tampering	A modified request changes the price paid	Server-authoritative pricing: the server works out every price	TheSkinProof
Regional payments and couriers	Card-only checkout loses local buyers; couriers are booked by hand	bKash, Nagad, SSLCommerz and cash on delivery; Pathao and Steadfast courier integrations	TheSkinProof: 4 payment rails. Sundor Skin: Pathao and Steadfast adapters
B2B tier pricing and credit	Buyers see the wrong price or order past their limit	Bronze, Silver, Gold and Platinum tiers with per-buyer overrides; credit limits and terms enforced in the database	Sundor Skin

Seller and buyer isolation

In a marketplace, the most expensive bug is one seller seeing another seller's orders, customers or payouts. Hiding a button in the interface is not security. The rule has to live on the server, and ideally in the database, so that a forgotten check in one screen cannot leak data.

On Sundor Skin, buyer data sits behind PostgreSQL row-level security, and 88 permission codes spread across 12 staff roles decide who can do what. On TheSkinProof, each audience (store, admin, seller, warehouse, auth) has its own portal. The budget side is in the cost to build a marketplace platform.

Authenticity and verification

On TheSkinProof, a product passes four gates before it is sold: invoice, physical, ingredient and batch code. If authenticity matters in your category, treat verification as a workflow with evidence and a decision, not as a checkbox a seller ticks. The story behind it is in building TheSkinProof, a verified marketplace.

Cash-on-delivery fraud

Where cash on delivery is normal, a refused parcel costs you the courier trip and ties up stock that someone else could have bought. TheSkinProof risk-assesses cash-on-delivery orders before accepting them, so the decision is made at the start of the order, not after a parcel comes back.

Stock correctness

Overselling happens when the stock check and the stock decrement are separate steps and two orders arrive between them. On TheSkinProof, per-variant inventory is decremented inside the same database transaction that creates the order, so the last unit can only be sold once. Sundor Skin adds batch-level stock: each order is allocated from the batch with the earliest expiry first (FEFO), then picked, packed and dispatched.

Regional payments and couriers

TheSkinProof runs 4 payment rails: bKash, Nagad, SSLCommerz and cash on delivery. On Sundor Skin, bKash and Nagad receipts are verified by staff. Both platforms book couriers through Pathao and Steadfast; on Sundor Skin these sit behind adapters, so a new courier is a new adapter rather than a rewrite. For US, UK and EU stores the rails differ but the principle does not: put each provider behind its own integration and decide up front what happens when it fails. Checkout design is covered in ecommerce checkout optimization.

B2B tier pricing, credit and approvals

B2B ecommerce is not B2C with a login. On Sundor Skin, a business applies, staff approve it, and the buyer receives a single-use invitation. Each buyer sits on a Bronze, Silver, Gold or Platinum tier, with per-buyer price overrides. Credit limits and terms are enforced in the database, not in the browser.

One small rule shows the thinking: a new delivery address waits 24 hours before credit orders can use it. That closes an obvious abuse path, where someone with a buyer's login sends goods on credit to an address the business has never seen. Invoices carry a QR code for verification, and changes are recorded in a hash-chained audit log. More on this in B2B ecommerce platform development.

Should you go headless?

Go headless when you need a custom, fast front end on top of an existing commerce engine, or several front ends sharing one backend. For a single, simple store it is usually extra moving parts without extra revenue.

Headless commerce means the storefront talks to the commerce backend only through APIs, so either side can change without breaking the other. TheSkinProof and Sundor Skin go one step further: they are fully custom platforms with their own backends, not headless front ends on a hosted engine. If a hosted engine is enough for you, Shopify vs custom e-commerce sets out when to stay on one.

In what order should you build a marketplace?

Money and stock rules first, storefront polish last. The rules are where mistakes cost most, and they are the hardest part to change once real orders exist.

1. Roles and isolation. Decide who can see and change what (buyer, seller, staff, warehouse) and enforce it on the server.
2. Catalogue, stock and pricing. Per-variant or per-batch stock, and a server that calculates every price.
3. Checkout and payments. One integration per provider, each with a defined failure path.
4. Fulfilment. Courier bookings, returns and refunds.
5. Storefront and growth features. Search, reviews, content and marketing tools on top of rules that already hold.

Tests are part of the build, not a phase after it. TheSkinProof runs 750+ automated tests and Sundor Skin runs 530+, gated in CI, so a change that breaks checkout, pricing or stock does not merge. The method is in how RAITHub tests: QA-first development, and RAITHub's phases are in how RAITHub delivers software.

How long does it take to build a marketplace?

A tightly scoped first release takes about 4–6 weeks; order, stock, pricing or payment logic behind an existing storefront takes about 6–12 weeks. A full marketplace keeps growing after launch, so plan for ongoing development rather than a finish line.

Engagement	Fits when	Typical length
SaaS & MVP (fixed scope, fixed price)	A first marketplace or B2B ordering release with a defined feature list	4–6 weeks
Backend & API (fixed scope, fixed price)	Order, stock, pricing or payment logic behind an existing storefront	6–12 weeks
Dedicated team (monthly)	Ongoing development after launch, where scope evolves	Month-to-month
Code Rescue	An existing store that is slow, fragile or losing orders	2-week diagnostic, then 2–4 weeks
QA & Reliability	You want CI gates and monitoring around a store you already run	Monthly retainer

To be plain about scale: TheSkinProof is 225K+ lines of code, live since Q2 2025 and still in active development. That is not a 6-week project. A 6-week project is a first release that does one job properly and can grow.

What it costs to build a marketplace or B2B store in 2026

Published agency guides put a basic multi-vendor marketplace MVP at roughly $40,000–$80,000 and a B2B marketplace at $120,000–$300,000 or more. Those are market figures, not RAITHub quotes. RAITHub publishes no rates; it gives a written fixed-scope estimate after a free 15-minute technical audit.

In commerce, the money goes on the rules the server has to enforce, not on the number of storefront pages. These are the lines that move a marketplace or B2B estimate most:

• Multi-vendor payouts. One checkout split between several sellers, commission held back, refunds charged to the right seller, and each seller paid out on schedule. That is a small ledger, not a checkout setting. Aalpha's July 2026 cost guide prices a single payment gateway at $5,000–$10,000 and split-payment logic for vendor payouts at $10,000–$20,000.
• Courier integrations. Each courier needs booking, tracking, returns and, where cash on delivery is normal, matching the cash the courier remits against the orders it delivered. The same guide puts shipping and logistics API integration at $3,000–$8,000.
• Roles and portals. TheSkinProof has 5; every extra audience needs its own screens and its own permission tests.
• Stock model. Per-variant stock is the simpler case. Batch-level stock with expiry dates and earliest-expiry-first allocation, as on Sundor Skin, adds receiving, picking and write-off flows.
• Trade pricing and credit. Tier prices, per-buyer overrides and credit limits enforced in the database are server work, not theme work.
• Verification. A multi-gate check such as TheSkinProof's four gates is a workflow with evidence, reviewers and states.

Scope	Typical market range (2026)	What drives it
Basic multi-vendor MVP	$40,000–$80,000	Seller onboarding, catalogue, one payment gateway, simple order flow
Mid-tier marketplace	$120,000–$180,000	Vendor payouts, several payment methods, courier integrations, seller reporting
B2B marketplace	$120,000–$300,000+	Buyer approval, tier pricing, credit terms, invoicing, integrations with the buyer's or seller's systems
Full-scale enterprise platform	$250,000+	Many portals and markets, high order volume, several payment and courier providers

Ranges from Aalpha's multi-vendor marketplace cost guide. Rates explain much of the spread between guides: Clutch's September 2026 software development pricing guide reports that most companies listed there charge $24–$49 an hour, with US-based firms typically at $50–$99. Budget for running costs too; Aalpha puts infrastructure for a small marketplace MVP at $300–$800 a month.

RAITHub's estimate names each of these lines separately, with its assumptions written down, so you can see which rule moves the number and cut scope knowingly. A first release is priced at fixed scope; growth after launch runs on a dedicated team, month-to-month. More market detail is in the cost to build a marketplace platform and the cost to build an ecommerce website, and how quotes work is on the pricing page.

Who is custom ecommerce development not for?

It is not for stores whose needs a hosted platform already meets, or for businesses that have to be selling within days.

• You sell one brand's products at standard prices. Use Shopify or a similar platform. You will launch sooner and spend less.
• You have not proved demand yet. Validate on a hosted store first; build custom once you know which rules matter.
• You need to be live in days. A custom platform with verification, pricing and stock rules needs a spec and a test suite. That takes weeks.

Why RAITHub for marketplace and B2B commerce

Because its commerce evidence is two live platforms with published numbers, and one of them RAITHub runs day to day. RAITHub is a founder-led software studio, founded in 2024 in Dhaka, Bangladesh, working with clients worldwide.

• TheSkinProof, the founder's own marketplace. 217 API endpoints, 5 role-based portals, 4 payment rails, a four-gate verification pipeline and 750+ automated tests, live since Q2 2025. Because RAITHub operates it, the team lives with the payment failures, courier exceptions and support tickets after launch, not only the launch. It is the founder's venture, not a client reference, so weigh it as evidence of engineering.
• Sundor Skin, a client's B2B wholesale platform. 146 PostgreSQL tables, 77 route pages, 12 staff roles built from 88 permission codes, batch stock allocated earliest expiry first, database-enforced credit limits and 530+ automated tests, live since September 2026.
• The money and stock rules are tested, not assumed. Both suites are gated in CI, so a change that breaks pricing, stock or checkout cannot merge.
• Two ways to engage. A fixed-scope build for the first release, then a dedicated team if the platform keeps growing. You own the code through a present-assignment IP clause, and an NDA is signed before any detailed discussion.

When RAITHub isn't the right fit

• You want a Shopify theme or app. RAITHub's published commerce work is custom platforms. An agency that works inside the Shopify ecosystem every day will know its apps and limits better.
• You need the store run for you. Merchandising, marketing, customer service and seller recruitment are not RAITHub services; it builds and tests the software.
• You want developers placed inside your team under your own management. RAITHub offers fixed-scope builds and dedicated teams, not staff augmentation.
• You need delivery in a language other than English. Specs, demos and support run in English.
• You want to cut testing to lower the quote. A CI-gated test suite is included in every RAITHub engagement and is not optional.
• You need a SOC 2 or ISO 27001 certified vendor. RAITHub is not certified; see the security page.

If none of these apply and you want RAITHub to build it, the ecommerce service page sets out what is included, or book the free 15-minute technical audit with what you sell, who sells it, and how buyers pay.

Last reviewed: 28 September 2026.

Frequently asked questions

Is custom ecommerce better than Shopify?

Not for most stores. Shopify or a similar hosted platform suits a single brand selling standard products at standard prices. Custom is better when sellers, verification, trade pricing, credit or local payment and courier rules are central to the business.

Can RAITHub build a multi-vendor marketplace?

Yes. RAITHub built and runs TheSkinProof, the founder's own verified-skincare marketplace, live since Q2 2025, with 5 role-based portals, 217 API endpoints, 4 payment rails, a four-gate product verification pipeline and 750+ automated tests.

Does RAITHub build B2B ecommerce platforms?

Yes. RAITHub built Sundor Skin, a gated B2B wholesale platform deployed in September 2026, with buyer approval, Bronze to Platinum tier pricing, database-enforced credit limits, FEFO stock allocation and 530+ automated tests.

How long does it take to build a marketplace?

A scoped first release typically takes 4–6 weeks, or 6–12 weeks for order, stock and payment backends, both at fixed scope. Larger marketplaces keep developing after launch, usually on a month-to-month team.

How much does a multi-vendor marketplace cost in 2026?

Aalpha's July 2026 cost guide puts a basic multi-vendor marketplace MVP at $40,000–$80,000, a mid-tier marketplace at $120,000–$180,000 and a B2B marketplace at $120,000–$300,000 or more. These are market figures; RAITHub gives a written fixed-scope estimate after a free 15-minute audit.

What drives the cost of a custom marketplace?

Vendor payouts, the number of payment rails and courier integrations, roles and portals, the stock model, pricing and credit rules, and verification steps. RAITHub publishes no rates; after a free 15-minute audit you get a written estimate with its assumptions.

Can RAITHub integrate payment providers outside Bangladesh?

Yes. TheSkinProof uses bKash, Nagad, SSLCommerz and cash on delivery, and PropDesk, another platform RAITHub built, collects rent through Stripe. The payment rails are chosen per market during the architecture spec.

Who owns the code RAITHub writes for my store?

You do. Full IP is assigned to you through a present-assignment clause, and an NDA is signed before any detailed discussion of your product.