CLIENT: Kadakaranna Kaubad OÜ🇪🇪Estonia

Kirbu - Resale Marketplace and Trust and Safety Back Office

Second-Hand Fashion and Kids' Goods Marketplace for the Baltics

Designed a resale marketplace for second-hand fashion and kids' goods shipped locker to locker across Estonia, Latvia and Lithuania, and built a clickable back office for it: listing moderation with AI suggestions checked by people, counterfeit and recall checks, trader detection and buyer protection disputes.

Scope of Work
Product design, UI build and clickable demo
Industry
Marketplaces / Resale Marketplace
AIEnterprise
Compliance & Certifications
GDPREstonian Personal Data Protection ActDigital Services ActDAC7General Product Safety RegulationPSD2 SCAEstonian Law of Obligations ActEstonian Consumer Protection Act

Most of our work is under NDA. Per agreement, we can share the work but not the company details - all names and identifying information shown here have been changed.

Impact

5 Screens
Clickable end to end

Dashboard, moderation queue, listing review, trader detection and disputes with the dispute panel. Every one of the 20 listings in the moderation queue opens its own review page.

5 Checks
Behind every listing review

Prohibited item classifier, keyword matches, price against the typical price, duplicate photos across accounts and an EU Safety Gate recall lookup, shown next to the AI suggestions and the seller's final values.

6 Reasons
Statement of reasons templates

Each removal names its ground: trademark law for counterfeits, the General Product Safety Regulation for recalls, DSA trader traceability, or the clause of Kirbu's terms, with the statement previewed before it is sent.

20 Listings
Realistic demo data

20 queued listings, 18 members across Estonia, Latvia and Lithuania, 14 trader cases, 12 disputes, 4 parcel locker carriers and 7 staff users.

The Problem

Kadakaranna Kaubad started as Tallinn parents running three large Facebook groups for children's clothes, prams and school uniforms. Buyers paid strangers by bank transfer, a fifth of the moderators' time went on "I paid and nothing came" posts nobody could resolve, fake designer bags and recalled cots appeared weekly, and some "private" sellers were shops clearing stock. At 90,000 members across the Baltics, the moderators could no longer keep up.

Our Solution

Kirbu puts every sale on one record: the listing and its photos, messages and offers, the payment held until the buyer confirms receipt, the locker label and tracking, and the evidence if something goes wrong. Buyer protection gives buyers a reason to pay a stranger, and the back office gives the team queues for moderation, trader detection and disputes instead of comment threads.

The Challenge: Strangers, Second-Hand Goods and EU Platform Rules

Kirbu is a small Estonian start-up, but as an online platform selling across three countries it carries the duties of the Digital Services Act, EU tax reporting and consumer law, while most of its sellers are parents clearing out a wardrobe.

Paying a Stranger

Sales between private individuals are generally not covered by consumer contract law, so buyer protection is Kadakaranna's own promise, not a legal right. Payment has to be held until the buyer confirms receipt, and every claim needs evidence both sides can see.

Counterfeits and Recalled Kids' Goods

Fake designer bags and recalled cots reached the old groups every week. High-risk brands need an authenticity photo set, used car seats, cots and cot mattresses are restricted categories, and reported kids' items are checked against EU Safety Gate alerts.

Private Seller or Trader

A person who sells regularly for profit may be a trader even if they call themselves private, and a trader's buyers generally get a 14-day right of withdrawal. Under the DSA traders must be traceable before they list, but Kirbu records declarations and evidence rather than making the legal call.

Platform Duties for a Small Team

Notice and action, a statement of reasons for every removal or restriction, appeals reviewed by someone other than the original decider, submissions to the DSA Transparency Database, and DAC7 reporting to the Estonian Tax and Customs Board for sellers above the de minimis threshold.

Our Approach: From Marketplace Rules to a Clickable Back Office

01

Spec and Rules Research

Mapped 8 roles, from buyer and private seller to moderator, finance and admin, and 52 screens: 24 mobile-first member screens and 28 back office screens. Modelled listings, offers, orders, shipments, claims, the wallet ledger, notices, decisions, appeals, trader signals and DAC7 records. Wrote down the rules the design follows: GDPR, the DSA, consumer law for trader sales, DAC7, PSD2 SCA, product safety, and Kadakaranna's own policies on fees, offers, shipping and withdrawals.

02

Clickable Demo on a Production UI Kit

Built the operator back office's 5 key screens on a production dashboard kit: the dashboard with KPIs, queues and carrier health, the moderation queue with bulk approve for low-risk triggers only, the listing review with its decision form, trader detection with tab actions that move a seller between stages, and disputes with a decision panel. Decisions move rows between tabs on local state.

03

What a Production Build Adds

The 24 member app screens and the remaining back office screens (notices, members, appeals, orders, refunds, wallet ledger, payouts, DAC7 reporting and audit log). Integration with a marketplace payment provider such as Stripe Connect or Adyen for Platforms, Omniva, DPD Pickup, Smartpost and LP Express, identity verification such as Veriff, Baltic business registers, VIES, the photo classification model, EXIF stripping, the DSA Transparency Database API and the DAC7 export. Staff SSO with two-step sign-in and logged personal data views.

Why It Works: People Decide, With the Evidence on Screen

AI Suggests, Moderators Decide

The photo model's suggestion and the seller's final value are both stored and shown side by side, with changes highlighted: a bag the model read as "Unbranded" at 87% that the seller listed as a high-risk designer brand stands out at once. Suggestions never decide moderation on their own.

One Screen per Listing Decision

Seller photos sit next to the brand's required reference set, with the automatic checks, seller account age and prior removals, earlier decisions and related notices. The moderator picks an action and reason code, sees the statement of reasons before it is sent, and the decision is recorded with an audit line.

Trader Status From Evidence

Each case shows a signal score with concrete examples, such as 131 sales in 83 days or the same jacket in 6 sizes, plus declaration history and register, VIES and identity results. Staff ask the seller to confirm, and new listings pause only after the deadline passes.

Disputes With a Clock and a Cost Bearer

A claim's age turns red once it passes the 3 working day target. The panel shows the listing as sold next to the buyer's claim, the seller's response and carrier tracking, and the decision sets the outcome, any partial amount and who bears the cost: seller balance, Kirbu goodwill or a carrier claim.

Related case studies

View all: Marketplaces
Logo
Oceanlab

Enterprise Software Engineering - Human + AI

Languages

Locations

Global Headquarters

21 Merchant Road, 3rd Floor, Singapore 058267

Main corporate office and leadership operations

North America Office

548 Market Street, PMB 12345, San Francisco, CA 94104

Client operations and business development

India Office

402 Shivalik Shilp II, SG Highway, Ahmedabad 380015

Engineering and delivery center

© 2026 - Copyright. All Rights reserved Oceanlab