Catawiki’s product is not its auction software. Its product is a few hundred subject-matter experts who decide what gets listed. Every lot submitted by a private seller is reviewed by someone who knows the category, and roughly speaking most submissions are rejected.
That rejection rate is the entire value proposition. Buyers bid confidently because someone competent already checked. If you want to build something similar, the interesting engineering is in the vetting workflow, not the bid button.
1. The submission and triage pipeline
This is the core system. A private seller submits an object; the platform must route it to the right expert, capture a decision, and communicate the outcome — at volume, quickly, without experts becoming a bottleneck.
What that requires:
- Category-aware submission forms. A watch needs reference number, movement, box-and-papers status. A painting needs artist, medium, dimensions, provenance. Generic forms produce unusable submissions.
- Enforced image standards. Required angles per category, minimum resolution, specific detail shots — hallmarks, signatures, serial numbers. Most rejections are for inadequate photography, so guiding this upfront saves expert time.
- Automated pre-screening. Filter obvious rejects before an expert sees them — missing required fields, prohibited categories, images below threshold. This is where you protect your most expensive resource.
- Expert routing with load balancing. Match submissions to specialists by category, availability and backlog.
- Structured decisions. Accept, reject with a coded reason, or request more information. The reason codes matter — they feed seller education and let you spot patterns.
2. Expert management is an operations problem
Your experts are the constraint on how fast the marketplace can grow, and they are usually contractors rather than employees. You need tooling for it:
- Throughput and decision-quality metrics per expert
- Consistency monitoring — two experts in the same category should reach similar decisions on similar objects
- Escalation for disputed or borderline lots
- Conflict-of-interest controls, since experts are often active dealers in their category
- Compensation tracking, typically per-decision or per-lot-sold
That last point deserves care. An expert paid per accepted lot has an incentive to accept. An expert paid per review has an incentive to review fast. Neither is neutral, and the incentive structure will shape your catalogue quality more than any policy document.
3. Curated scheduling instead of open listing
Catawiki groups accepted lots into themed weekly auctions rather than listing them the moment they are approved. This is a deliberate choice with real benefits: it concentrates buyer attention, allows category marketing, and creates a recurring event bidders plan around.
The trade-off is seller patience — an object may wait weeks for the right sale. Your platform needs a scheduling layer that assigns approved lots to upcoming themed auctions, with capacity limits per sale so no single auction becomes an undifferentiated dump.
4. The bidding engine (the easy part)
Standard timed auctions with proxy bidding, tiered increments, soft-close anti-sniping and outbid notifications. This is well-understood, well-solved territory — see proxy bidding for the mechanics.
Do not build this from scratch. License it, and put your engineering budget into the vetting pipeline where it actually differentiates.
5. Payments, escrow and cross-border complexity
A curated marketplace with private sellers has payment requirements a single-seller auction house does not:
- Buyer funds held until delivery is confirmed, then released to the seller
- Seller payouts across many countries and currencies
- Marketplace tax obligations that vary by jurisdiction and have tightened considerably
- Customs documentation for cross-border shipping
- Category-specific restrictions — ivory, certain weapons, cultural property with export controls
That last one is not optional in a collectibles marketplace. Cultural property law is genuinely complex and the penalties are serious. Build a category rules engine that blocks prohibited listings by jurisdiction at submission time.
6. The economics of curation
Be clear-eyed about this. Expert review costs real money per submission, and you pay it on rejections too. If your average accepted lot sells for $200 and your commission is 12%, you earn $24 — while possibly having paid for four reviews to get there.
Which means curated marketplaces need either a high average lot value, a high acceptance rate, or very efficient pre-screening. Most successful ones have all three, achieved by pushing hard on automated filtering and seller education so experts only see plausible submissions.
The strategic version of this trade-off is covered in our comparison of curated and aggregated platforms. Budget guidance is in marketplace development costs, and the platform layer in multi-vendor marketplace platform.
Frequently asked questions
Can I automate expert review with AI?
Partially, and it is genuinely useful for pre-screening — flagging missing data, poor images, obvious category errors and likely reproductions. Final authentication in high-value categories still needs a human, and buyers know the difference.
How many experts do I need to launch?
Start with one deep category and one or two genuine specialists. A curated marketplace that is excellent in one vertical beats one that is mediocre across ten.
What is the hardest part of this build?
The submission pipeline and expert tooling, by a wide margin. The auction mechanics are commodity. Talk to our team about layering curation onto a proven engine.
