Skip to content
Kreezalid

Getting started

How to Create an Online Marketplace in 2026: From Idea to First Transaction

How to create an online marketplace: prove one transaction, win the hard side, design the payments, then pick no-code, WordPress or custom code.

Kreezalidupdated 18 min read

How to create a marketplace, step by step

To create an online marketplace, run one transaction by hand with a narrow group, recruit the harder side, design how the money moves, and only then choose how to build. For most first launches the route is a no-code platform; a WordPress plugin stack fits if you already run a store; custom code pays off only when the workflow itself is your edge. Custom development is the slow route: one agency puts a marketplace app at 3-5 months to a year or more.

Demand for online buying is not the constraint. In the EU in 2025, 78% of internet users had bought or ordered goods or services online, according to Eurostat. In the same survey 63% of online buyers reported no problem, while 20% hit slow delivery, 12% a difficult website and 10% a wrong or damaged item.

A marketplace inherits those delivery problems from its sellers and takes the blame for them, which is why the order of the steps matters more than the feature list.

The short answer: how do you create an online marketplace?

  1. Define the transaction: what is exchanged, by whom, and what you earn when it happens.
  2. Validate by running one transaction by hand, with real buyers and a real seller.
  3. Win the hard side, inside one category or one city.
  4. Choose the build route: no-code platform, WordPress or custom code.
  5. Design the money flow: who holds funds, when sellers are paid, who bears refunds.
  6. Build trust into the transaction, before, during and after payment.
  7. Scope an MVP that completes the whole transaction and nothing more.
  8. Launch in sequence and track liquidity before you widen anything.

The order exists because each step fixes a constraint for the next: the test shows which side is hard, and the route you pick has to carry the money flow you designed.

1. Define the transaction before choosing features

Write one sentence before opening any software: [Buyer segment] books, buys or contacts [seller segment] for [specific outcome] in [category or place], and the marketplace earns [fee] when [value event] happens. If you cannot fill the brackets, you are not ready to choose features, because the brackets are the features.

Match the marketplace model to the workflow

The model sets what the first version must do. This table is the working version; each model name links to a page that goes deeper.

Model Core transaction What the MVP needs Natural revenue model Main operational risk
Services A buyer books or hires a provider Profiles, availability or quote requests, messaging, payment, reviews Commission, lead fee, subscription No-shows, scope disputes, deals moved off-platform
Rentals A buyer reserves an asset for a period Calendar, duration pricing, deposit, handover rules Commission, plus a service or protection fee Double bookings, damage, deposit disputes
Products A buyer purchases from one or more vendors Catalogue, stock, cart, shipping, returns, split payouts Commission, listing fee, seller subscription Fulfilment, returns, unsafe or counterfeit goods
Directory A buyer finds and contacts a lister Listings, search, lead form, paid plans, featured placement Subscription, listing fee, lead fee Weak lead quality, low renewal value
Peer-to-peer Two individuals trade with each other Profiles, messaging, reviews, payment Commission, listing fee Fraud, disputes between people with no relationship
B2B A company buys from or rents from another Business accounts, quotes, invoicing, repeat orders Commission, subscription Long cycles, negotiated prices, deals closed offline

B2B is a layer over any of the other models: a wholesale shop is a product marketplace with business accounts on top.

Choose monetisation with unit economics, not imitation

A competitor’s take rate is not yours. What you can charge depends on the value you create and the costs you absorb. Four equations are enough:

  • GMV = completed transactions × average order value
  • Marketplace revenue = GMV × take rate + subscription, listing and lead revenue
  • Contribution margin = revenue − payment fees − refunds − incentives − variable support cost
  • Seller net earnings = GMV − marketplace fees − seller-borne payment or fulfilment costs

GMV and revenue are not interchangeable: processing fees, refunds and support come out of the slice you keep, not out of GMV.

Charge for the value event you can observe and defend. If the payment stays on the platform, a commission usually lines up with what sellers gain. If deals close offline or carry high values over long cycles, a subscription or a qualified-lead fee is easier to enforce. Kreezalid itself takes no commission on sales: the operator sets and keeps its own, and pays a flat monthly plan. The business model guide goes through the options.

2. Validate by running one transaction by hand

You know there is demand when a stranger pays you for a transaction you arranged with a spreadsheet and a payment link. A waitlist does not tell you that.

Interview both sides

Talk to sellers and buyers separately, and ask about the last real transaction, not your idea:

  • How did the buyer find a provider, and where did trust break down?
  • How was the price agreed, and how was payment made?
  • What went wrong afterwards, and what did the seller pay to win that customer?

Find people where they already complain about the problem, such as trade forums or professional associations. Friends and family give kinder answers than useful ones.

Prove the transaction manually

Act as the marketplace. Recruit a few relevant sellers, turn their offers into consistent listings, find real buyers by direct outreach, match the two sides yourself and note every objection and failure point.

One example from the sources we collected is Cury, a marketplace now doing more than 25 million in annual GMV, according to Sharetribe’s podcast. It started as dress rental inside one sorority, with payments by Venmo and the rentals run manually. Not a city or a university: one sorority.

Casey Winters, quoted on the same podcast, puts the goal this way: get an initial network effect to work, which he calls liquidity or product-market fit, and do not expand to new markets before it does.

A validation checklist

Treat these as gates, not as a score.

  • Repeat transaction. Several manual matches reached a satisfactory outcome for both sides, and at least some came back.
  • Paid. A booking, order or paid pilot with a real budget, not an email address.
  • Both sides came without being pushed. If you had to persuade every party from scratch, the offer or the supply is the problem, not the software.
  • Known automation list. You can say which steps you repeated by hand and would automate first.

If you cannot complete the transaction manually, software will not repair the market design.

Launch readiness is not product-market fit

A few paid transactions justify the next experiment. They do not show product-market fit. After launch, look for buyers returning when the need recurs, and for a repeat window that matches the transaction: a weekly service and an annual equipment rental should not share a 30-day retention target.

3. Win the hard side with density, not breadth

Equal numbers of buyers and sellers solve nothing. One small market where a buyer finds what they need solves the cold start.

Identify the hard side

The hard side is the one that is scarcer, more selective or more valuable to the other: specialist clinicians, licensed tradespeople, event venues, vintage dealers. In some B2B models it is a committed buyer whose orders attract suppliers. Secure it first and use it to pull in the other.

Start with a wedge

The podcast host’s advice is blunt: start small or constrain your marketplace. In practice you constrain one of five things: the city or service radius, the category, the buyer segment, the time window, or the job to be done.

Andrew Blachman of Marketplace Capital, on another Sharetribe podcast episode, says it is far better to be number one than number two or three. A narrow wedge is how a newcomer gets to be number one somewhere.

Seed supply by hand, without fake accounts

Early supply is operations work: build the first listings with sellers, standardise service descriptions, waive fees for a founding cohort.

One vendor, iScripts, advises creating a few user accounts and sample products so the site does not look empty. We would not. This is an opinion, not a legal point: your first real buyer is trusting that a seller exists, and that trust is the asset you are trying to build. A few real sellers with complete listings beat a full catalogue that nobody can book.

Measure liquidity every week

Liquidity is whether a qualified buyer finds a suitable seller in an acceptable time. Two numbers carry most of it:

  • Match rate: requests that end in an accepted match, divided by qualified requests.
  • Time to match: the median time from request to accepted match.

Add completion and repeat rate once you have the volume. Fix the definitions once and keep them stable: a rising registration count can hide falling liquidity.

More on this stage: three ideas to solve the chicken-and-egg problem, who to target first and how to onboard sellers.

4. Choose how to build: no-code platform, WordPress or custom code

Most guides start here. It comes fourth because the earlier steps tell you what you need to buy, and sometimes that you should not buy yet.

The three routes compared

Every time claim in this table comes from the vendor who makes it. None is a measurement.

Route Time to a usable first version Upkeep Payments Best fit What it does not do
No-code platform Kreezalid promises about 7 days; Yo!Kart says 5-7 working days; Marketplacer says “within weeks” for an existing retailer Depends on the vendor Depends on the vendor Standard marketplace workflows, first launches Workflows the platform cannot be configured to follow
WordPress plus a multi-vendor plugin Depends on the plugins and on who assembles them Shared across WordPress, the theme, each plugin and the payment add-ons Often assembled from add-ons A store you already run, to which you add sellers Hands-off upkeep: every update is yours to test
Custom code 3-5 months to a year or more, per Codica Yours: hosting, security patches, payment-provider changes Designed from scratch A workflow no platform can follow A fast, cheap first test

On the WordPress route, Dokan claims more than 50,000 marketplace owners, and Webkul, a WooCommerce agency, says admins can set a global commission rate along with seller-specific rates. Both are vendor statements. The WordPress route is worth weighing honestly before you commit.

Kreezalid’s 7-day figure is a product promise, not a guarantee. Content, supplier recruitment and legal review set the real launch date.

Without code, model by model

Kreezalid covers services, rental, products, directory, peer-to-peer and B2B. Whichever platform you look at, ask for a demo of your own model and check these things:

Model What to check in a demo When a platform is the wrong tool
Services Availability calendar or quote requests, how a booking becomes a payment, how reviews are tied to a completed job The price is negotiated line by line and the work is delivered off-platform
Rental Availability by date, duration pricing, deposit and partial refund, damage handling Rentals depend on logistics or insurance a standard flow cannot express
Products Split cart across vendors, stock, shipping per seller, returns Order handling must sync live with a warehouse or ERP system
Directory Listing plans, featured placement, lead form, renewals The directory is a data product with its own feed and rules
Peer-to-peer Profiles, messaging, reviews, how a dispute between two individuals is handled Your trust model needs checks the platform has no place for
B2B Business accounts, quotes, repeat orders, approval steps Pricing, catalogues and approvals live in your ERP, PIM, CRM or single sign-on

Auctions and reverse marketplaces are not in Kreezalid’s list of marketplace types. If your model is either of those, check before you commit to any platform, ours included.

What a platform costs, and what a custom build costs

Kreezalid’s monthly plans are flat, with no commission on sales: Build is €39, Launch €129, Grow €299 and Extend €599. Plus and Enterprise are on quote. The prices are in euros and we have not converted them. Kreezalid offers more than 50 templates, and 14 house themes are sold at €179 to €299 excluding VAT, as a one-time payment with updates included. The pricing page shows what each plan includes.

A custom build cannot be priced from the sources collected for this article. The only figure we found, from Codica, is $20,000 to $200,000 for a cross-platform app, which is not specific to marketplaces, so do not use it as an estimate. The cost guide covers the cost side in more detail.

Compare total cost of ownership, not the first invoice.

When not to buy a platform yet

Buying software before you have the evidence is the most common wasted spend. Not yet:

  • You have no manual transaction behind you yet. Go back to step 2. A configured platform tells you nothing about demand.

Not this kind of platform, at all, if:

  • The workflow is your differentiator and cannot be configured. That is the case for custom code.
  • You need auctions or a reverse marketplace, which are not in Kreezalid’s list of types.
  • You already run a WooCommerce or similar store and only want sellers on it. A plugin on the store you have may be enough.
  • The integration is the project: ERP, PIM, CRM or single sign-on. The Enterprise page covers marketplace projects integrated with ERP, PIM, CRM and SSO, on quote.

If the transaction is proven and none of these applies, the templates give you a starting point by model, and the pricing page shows the plans. The trial lasts 14 days and needs no credit card. To compare further, see the no-code marketplace guide, the marketplace builder page, Kreezalid against Sharetribe and a list of platforms.

5. Design the money flow before the checkout

A marketplace payment involves a buyer, one or more sellers, you and a payment provider. Settle who does what before you pick any checkout setting, because it drives refunds, payouts and reporting.

Questions to settle first

  1. Who is the merchant of record?
  2. When is the buyer charged or authorised?
  3. Is the money split at payment or transferred later?
  4. When does a seller become eligible for a payout?
  5. Who pays processing fees?
  6. Who absorbs refunds, chargebacks and negative balances?
  7. Do you need deposits, partial refunds or milestone payments?
  8. Which countries, currencies and payment methods do you need?
  9. What identity and business data must you collect from sellers?
  10. What transaction records must you keep, and for how long?

What each model needs

Qualitative, from experience rather than a benchmark:

Requirement Services Rentals Products Directory
Seller onboarding with identity checks Usually Usually Usually Only if you take payment
Commission split Usually Usually Usually Optional
Scheduled payouts Usually Usually Usually Optional
Deposit or pre-authorisation Sometimes Often Rarely No
Partial refund Often Often Often Rarely
Multi-vendor cart Rarely Rarely Sometimes No
Recurring billing Optional Optional Optional Often

Providers

Kreezalid has Stripe Connect and Lemonway built in. If you need another provider, ask the Kreezalid team during the demo before you commit. We make no claim here about which countries each provider covers for sellers, so check that against where your sellers are.

Mangopay describes itself as a licensed Electronic Money Institution for platforms and marketplaces, with custom, volume-based pricing and no fixed public tiers. That detail is why a licensed provider matters: do not route seller money through an ordinary business bank account. An adviser can tell you what applies in your markets.

The word “escrow” has a specific legal meaning in many jurisdictions, so we do not use it for any Kreezalid feature. The payment solutions guide and the payments page show how the pieces fit together.

6. Build trust into the transaction

Trust is the system that reduces uncertainty before, during and after payment. A testimonials section is a small part of it. The Eurostat survey cited at the top found 4% of online buyers reporting fraud.

Before, during and after

Before payment. Verify seller identity and, where relevant, licences or insurance. Require complete, structured listings, and show the total price, fees and cancellation terms before the buyer pays.

During the transaction. Keep messaging on the platform so agreements are on record, confirm each milestone, and give both sides a visible way to reach support.

After the transaction. Collect reviews tied to completed transactions, publish a simple refund and dispute process with evidence requirements and deadlines, and feed dispute causes back into your listing rules.

The longer treatment is in the marketplace trust guide and in the article on protecting users against fraud.

Compliance is part of the product

Tax reporting, consumer protection and product safety obligations depend on your country, your role and your model. Take advice before launch rather than after the first complaint. We name no specific regulation because they differ too much by market.

7. Scope an MVP that completes the full transaction

A first version is the smallest system in which a real buyer and a real seller finish the exchange safely.

Features nearly every transactional marketplace needs

  • Separate buyer, seller and admin roles, with seller onboarding and profiles
  • Listing creation and moderation, search and filters
  • An enquiry, booking or cart flow, with a payment and payout flow built for marketplaces
  • Notifications, cancellations, refunds and dispute intake
  • Reviews tied to completed transactions, and admin visibility into users, listings and transactions

Add only what the model requires

Services need calendars and quotes. Rentals need availability, deposits and damage evidence. Products need stock, shipping and returns. Directories need membership plans and lead forms.

Delay what does not unlock the first transaction

Native apps, recommendation engines, loyalty schemes and seller tiers can wait, unless your pilot showed they decide whether a transaction happens.

The pages a marketplace website needs

Most marketplaces need the same set: a homepage that routes each side, search results, listing pages, seller profiles, checkout or booking, dashboards for both sides, a “how it works” page, terms and refund policies, and a help page. Put your own domain on it: it carries the trust and the search visibility you earn.

Turn the manual pilot into an acceptance test

The run you did in step 2 becomes your test script. Use the seller’s real offer to define listing fields, the buyer’s selection criteria to choose filters, and the agreed fee to check the payment flow. Then replay the journey as buyer, seller and operator, including a cancellation and a refund.

If you set up on Kreezalid, every marketplace has an MCP server that lets Claude, ChatGPT, Claude Code, Cursor or VS Code set up categories, listing attributes, order types, pages and the blog. It never touches listings, orders, members, messages, reviews or payment configuration. It is included on every plan and stays off until an administrator switches it on.

The feature test

When a feature request arrives, ask four questions in order:

  1. Does it help the chosen buyer and seller complete the core transaction?
  2. Does it improve liquidity, payment reliability, trust or compliance?
  3. Did the problem show up in real transactions?
  4. Can it be handled manually until volume justifies automating it?

If the answer to the first three is no, it does not belong in the MVP. The MVP marketplace guide goes further on scoping.

8. Launch in the right sequence

Technical setup and a credible market launch are different projects. Seven days is Kreezalid’s product promise for the configuration; supplier recruitment, legal review and content decide the real date. The day-by-day split below is a suggested schedule, not a Kreezalid commitment.

Before the build

Complete the manual matches, fix the fee and the failure policy, and recruit the founding supply cohort.

Days 1 to 2: configure the model

Pick the template closest to your model, set user roles, listing fields, categories and geography, and apply your brand.

Days 3 to 5: configure the transaction

Connect the payment provider, set commissions, payout rules, refunds and deposits, write the policies, and add the first complete listings.

Days 6 to 7: test and soft launch

Run test transactions for the happy path and the failures, then invite a small buyer cohort and support the first transactions by hand.

The first 30 days live

Fix the most frequent transaction failure each week, and widen the wedge only when matches repeat.

9. Track the metrics that show a marketplace is working

One dashboard with four groups is enough for the first 90 days.

Liquidity

Active supply, match rate, time to first response, time to match, and searches with no result.

Transaction quality

Checkout or booking completion, cancellation and refund rate, dispute rate and resolution time.

Economics and retention

Contribution margin per transaction, and repeat rate over a window that fits the transaction.

Acquisition

Cost per activated buyer and seller, and conversion from channel to first transaction.

The marketplace metrics guide defines each of these in detail.

Frequently asked questions

What is an online marketplace?

An online marketplace is a platform where independent sellers and buyers find each other and trade products, services, rentals or leads. The operator provides listings, search, payments and trust rules without necessarily owning any of the stock. The definition guide goes deeper.

What is the difference between a marketplace website and an e-commerce website?

An e-commerce site sells its own products. A marketplace connects many independent sellers with buyers and earns on the transactions between them, so it also needs seller onboarding, third-party listings, split payments and a way to settle disputes between users. The ten differences are set out in a separate article.

Can I create a marketplace without coding?

For services, rental, products, directory, peer-to-peer and B2B models, yes: Kreezalid covers all six, and the promise is about seven days to launch. Auctions and reverse marketplaces are not in its list, so check those in a demo before you commit. Custom code makes sense only when no platform can follow your workflow.

How much does it cost to create an online marketplace?

On Kreezalid, plans are flat monthly fees with no commission: €39, €129, €299 or €599, with Plus and Enterprise on quote. A custom build cannot be priced from sources we could verify; the one range found, $20,000 to $200,000, is for any cross-platform app. Our cost guide explains what to budget.

How long does it take to build an online marketplace?

Vendors give different ranges: Kreezalid about 7 days, Yo!Kart 5-7 working days, and Codica 3-5 months to a year or more for a custom app. All are claims by the seller of that route. Supplier recruitment, legal review and content usually take longer than the technical setup, whichever route you choose.

Should I develop a marketplace from scratch?

Only when the workflow itself is your edge and no platform can be configured to follow it. Otherwise you carry hosting, security, payment-provider changes and bug fixes for as long as the marketplace runs. The comparison with building from scratch sets out the costs and trade-offs of each route.

Which side of the marketplace should I acquire first?

The hard side: the scarcer, more selective or more valuable one. It is usually supply, but in some B2B models a committed buyer is what attracts suppliers. Sharetribe’s podcast advises constraining the marketplace, and Casey Winters says to reach liquidity before entering new markets.

How do online marketplaces make money?

The usual models are a commission on each transaction, a subscription, a listing fee and a lead fee, or a mix. Pick the one that charges for a value event you can observe. Kreezalid takes no commission: the operator sets and keeps its own, and Kreezalid charges a flat monthly plan.

Do I need escrow for a marketplace?

Not necessarily. “Escrow” has a specific legal meaning in many jurisdictions. What you need is a defined flow: when the buyer is charged, how funds are split, when sellers are paid and who bears refunds. Use a licensed provider, such as an electronic money institution, and take legal advice for your markets.

What features does a marketplace MVP need?

Buyer, seller and admin roles, seller onboarding, listings, search, a booking, cart or enquiry flow, a payment flow built for marketplaces, notifications, policies, reviews tied to completed transactions and admin oversight. Add calendars, deposits, stock or subscriptions only if your model requires them.

Where to go from here

Still validating. Go back to step 2 and run the transaction by hand. A marketplace business model is easier to choose after a real payment.

Transaction proven. Start the 14-day trial, no credit card needed, and pick a starting point from the templates or compare plans on the pricing page.

Not sure it fits. Book a 30-minute demo call on the office hours page and walk through the transaction you tested.