How to Build an App Like Fiverr

Productised gigs vs bidding

The first product decision, and it changes everything downstream.

Productised (Fiverr's model): sellers publish fixed-scope, fixed-price packages. Buyers browse and purchase like a product. Lower friction, easier search, better conversion — and it constrains what can be sold to work that packages cleanly.

Bidding (Upwork's model): buyers post a brief, sellers propose. Handles complex custom work, at the cost of much higher friction and a matching problem on both sides.

Productised is the better MVP choice for almost everyone. You can add briefs later; starting with bidding means solving matching before you have liquidity to match with.

Escrow and milestones are the product

Money flow is the core mechanic:

Buyer pays up front; funds are held by the platform
Seller delivers
Buyer accepts, or an auto-accept window expires
Funds release to the seller minus your fee

For larger engagements, milestone-based release is the same loop repeated. What you must design deliberately: the auto-accept window (buyers go quiet and sellers must still get paid), revision limits before a delivery counts as complete, and what happens when a buyer disappears mid-order.

Disputes will happen and need a defined path

Unlike goods, "was this delivered properly" is a judgment call. You need stated acceptance criteria captured at order time, a revision process before dispute is available, evidence attached to the order rather than in email, and a human decision-maker with a documented resolution policy.

At MVP scale the human is you, and that's fine — but the policy has to exist before the first dispute, not after.

Off-platform leakage is the existential threat

Once a buyer and seller work together successfully, both save money by going direct next time. Defences that actually work:

Escrow protection that genuinely benefits both sides
A fee proportionate to the value the platform adds
Contact-detail detection in messaging
Repeat-buyer incentives that make staying cheaper than leaving

A punitive-only approach fails; the platform has to remain the better option.

Cut for an MVP

Video calls, time tracking, team accounts, portfolio sites, and skill tests. Launch with one vertical of services so search and quality control are tractable.

Related builds

How to build a marketplace MVP — the general playbook
How to build an app like Airbnb — inventory rather than services
How to build a job board — hiring rather than transacting

FAQ

Should a services marketplace use fixed packages or bidding?

Fixed packages for an MVP. Productised gigs have far lower friction, simpler search, and better conversion, whereas bidding requires solving two-sided matching before you have the liquidity to match with. You can add briefs and proposals later once supply is dense.

How does escrow work on a freelance marketplace?

The buyer pays up front into platform-held funds, the seller delivers, and funds release on acceptance or when an auto-accept window expires. The details that need deliberate design are the auto-accept window, revision limits before delivery counts as complete, and what happens if a buyer disappears mid-order.

How do you stop freelancers and clients going direct?

Make the platform the better option rather than just penalising leaving: escrow that protects both sides, a fee proportionate to the value you add, contact-detail detection in messaging, and repeat-buyer incentives. Purely punitive approaches push the behaviour underground instead of preventing it.