‍How a Classified Marketplace, a Collectible Marketplace, and a Retail Marketplace, all scale Commerce Media with Topsort‍

Poshmark connects millions of individual sellers with fashion shoppers. 1stDibs helps buyers discover distinctive furniture, art, jewelry and collectibles. Falabella operates a broad marketplace spanning categories, advertisers and markets. All three use sponsored listings. But each creates a fundamentally different monetization challenge.

Poshmark needs to make long-tail, seller-generated inventory easy to advertise. 1stDibs needs to find relevant products across a highly differentiated catalog before an auction can even begin. Falabella needs to optimize thousands of different commercial environments through one system at scale.

Topsort provides the AI monetization infrastructure behind these experiences. Its technology brings together candidate retrieval, relevance, auctions, automated bidding, budget pacing, optimization and measurement—and adapts each layer to the marketplace using it.

Here is how that works in practice.

Poshmark: Monetizing Long-Tail Resale Inventory

Resale marketplaces create one of the hardest problems in marketplace advertising.

On Poshmark, millions of listings come from individual sellers. Many products are unique, available only briefly or have limited advertising history. Traditional ad systems that depend heavily on historical click-through or conversion data tend to favor listings that have already accumulated traffic. That leaves much of the marketplace difficult to monetize.

Topsort addresses this by learning from the marketplace’s broader commerce activity. Searches, clicks, purchases and other organic interactions help its models understand what shoppers respond to, even when an individual listing has little paid advertising history.

For Poshmark, Topsort developed custom machine-learning models that use these marketplace signals to improve prediction across sparse and fast-changing inventory.

Making advertising accessible to everyday sellers

Poshmark’s sellers are another important part of the challenge. Most are not professional media buyers. They should not need to manage keyword lists, CPCs and constant bid adjustments to promote their closets.

Topsort’s Bidless technology reduces that complexity.

A seller defines a budget and objective. Topsort evaluates each eligible opportunity using predicted outcomes such as click probability, conversion likelihood and expected value. The system then determines how aggressively the campaign should participate in each auction.

This gives sellers access to sophisticated performance advertising without requiring sophisticated ad operations.

Protecting relevance while increasing revenue

Automated bidding still has to protect the shopping experience. A larger budget should not allow an irrelevant listing to dominate a valuable placement.

Topsort combines commercial value with relevance and Quality Score signals. Ranking reflects both the value of the bid and the likelihood that a product belongs in that particular shopping experience.

Measurement also follows how resale shoppers actually behave. Someone may click a promoted item, visit the seller’s closet and purchase a different product. Topsort’s halo attribution can capture this seller-level impact instead of measuring only purchases of the exact advertised listing.

For Poshmark, marketplace monetization requires more than inserting ads into a feed. It requires making unique inventory predictable, advertising simple for individual sellers and results measurable across the complete shopping journey.

1stDibs: Finding the Right Sponsored Products Before the Auction

1stDibs presents a very different marketplace monetization problem.

Its catalog spans furniture, art, jewelry, fashion and collectibles. Products are highly differentiated, and shoppers discover them through search, category pages, recommendations, creators and product-detail pages.

In this environment, the first challenge is not choosing which advertiser wins. It is deciding which products should be considered at all.

Candidate retrieval for a discovery-led marketplace

Running every product through every auction would be slow and largely irrelevant. Topsort first retrieves a smaller set of plausible sponsored candidates from the wider catalog.

Candidate retrieval identifies products relevant enough to enter the auction. Auction ranking then determines which eligible candidate should win.

That distinction becomes especially important outside search.

A search query provides a direct signal of shopper intent. A product-detail page may have no query. Topsort can instead use product attributes, marketplace taxonomy, contextual signals and semantic similarity to identify sponsored products related to the item being viewed.

This allows 1stDibs to extend marketplace monetization beyond keyword search and into the discovery experiences that define its platform.

Combining marketplace intelligence with auction intelligence

1stDibs already has sophisticated systems for organizing products and understanding discovery. Topsort can incorporate marketplace-provided relevance signals alongside its own decisioning, preserving the marketplace’s product intelligence within the monetization process.

Once Topsort retrieves potential candidates, it evaluates campaign status, available budget, seller and product eligibility, placement rules and other marketplace-specific constraints.

Eligible candidates then enter an auction that can combine:

• Bid or predicted opportunity value
• Product and contextual relevance
• Quality Score
• Budget pacing
• Predicted performance
• Marketplace-provided signals

For 1stDibs, Topsort performs this decisioning within sub-50-millisecond auction latency.

The result is an architecture that moves from context to retrieval, eligibility, scoring, auction, pacing and winner selection in real time.

For a marketplace built around discovery, monetization begins by finding the right products before the auction starts.

Falabella: Optimizing Marketplace Advertising Across Categories at Scale

A large multi-category marketplace creates complexity through both scale and variation.

Electronics, beauty, fashion, home and other categories behave differently. They have different advertiser density, order values, conversion rates, buying cycles and levels of competition.

One fixed set of marketplace advertising rules cannot optimize all of them.

Real-time auctions for different commercial environments

Topsort runs auctions at the request level. Each advertising opportunity is evaluated against the products and advertisers eligible at that moment.

A competitive electronics auction can therefore behave differently from an auction in a lower-demand furniture category—without requiring Falabella to operate separate advertising systems or manually configure pricing for every environment.

In Falabella’s initial test, Topsort generated more advertising revenue from the same number of sponsored placements. The improvement came from better decisioning rather than adding more ad inventory.

That distinction matters. Marketplace monetization should improve the value of an existing customer experience, not simply increase ad density.

Balancing bids, relevance and marketplace yield

At Falabella’s scale, ranking products purely by bid would create significant risk. Irrelevant products could repeatedly win valuable placements, reducing shopper engagement and eventually hurting advertiser performance.

Topsort combines bids with relevance and Quality Score so each auction considers both immediate commercial value and shopper fit.

The system can optimize toward stronger click-through rates, conversion, advertiser returns and long-term marketplace yield.

Spending each budget at the right moment

Budget pacing adds another layer of intelligence.

Large advertisers may have campaigns that run for weeks or months across millions of possible auctions. Winning an auction is not always the best use of the advertiser’s remaining budget.

Topsort continuously considers:

• Remaining campaign budget
• Current and expected spend rate
• Forecast traffic
• Time left in the campaign
• The quality of the present opportunity
• Expected future opportunities

For advertisers using Bidless, Topsort can also estimate the value of each impression using predicted click-through rate, conversion probability, expected order value, budget and target return.

This allows the bid to adapt to each opportunity instead of relying on one static CPC.

Migrating an established marketplace advertising business

Falabella already operated an active retail media business. Modernizing its technology could not require starting over, interrupting active campaigns or discarding historical performance data.

Topsort supports the staged migration of campaign structures and historical signals, helping marketplaces preserve advertiser continuity and reduce model cold-start effects as they transition to a new monetization stack.

For Falabella, Topsort provides one infrastructure layer capable of making different decisions across categories, campaigns and markets at scale.

One AI Monetization Infrastructure Layer, Built for Different Marketplaces

Poshmark, 1stDibs and Falabella all offer sponsored listings. The technology required to make those listings successful is different for each marketplace. For Poshmark, Topsort makes sparse, seller-generated inventory predictable and gives individual sellers a simple way to advertise.

For 1stDibs, Topsort retrieves relevant candidates from millions of differentiated products before running the auction.For Falabella, Topsort optimizes relevance, bids, budgets and performance across categories and markets at scale.

The shared foundation is Topsort’s AI monetization infrastructure: a real-time decisioning system that brings together marketplace context, retrieval, eligibility, relevance, bidding, pacing, prediction and measurement. Each marketplace keeps its own identity, economics and customer experience. The infrastructure adapts around them.

Different marketplaces. Different monetization problems. One intelligent infrastructure layer.

The visible ad product may be the same: a sponsored ad. What changes is the intelligence underneath it.

For Poshmark, Topsort has to make sparse, seller-driven inventory predictable and easy to advertise. For 1stDibs, it has to retrieve relevant candidates from millions of unique products before auction ranking starts. For Falabella, it has to optimize bids, relevance, budgets, and performance across very different categories at scale.

The common layer is a real-time decisioning system that combines marketplace context, relevance, eligibility, bidding, pacing, prediction, and measurement differently depending on how the marketplace works.

That is why marketplaces across resale, B2B, delivery, travel retail and multi-category commerce are building with Topsort. That is what makes marketplace monetization fundamentally different.