The Full Platform Bill

Every serious seller knows the take rate. Amazon charges referral fees of roughly 8-15%, depending on the product category, plus Fulfillment by Amazon (FBA) fulfillment costs if you use their logistics network. These figures are clearly displayed in every seller dashboard, quoted in onboarding documentation, and incorporated into the initial spreadsheets built before launching a marketplace business.

While referral fees have remained relatively stable over the past decade, surrounding costs have steadily accumulated.

By 2022, Amazon retained over 50% of third-party seller revenue across all fees and advertising, a significant increase from approximately 33% in 2016. The referral fee component remained largely unchanged. The advertising component did not. This discrepancy is the core issue, yet the take rate chart doesn't reflect it.

The First Layer: The Fee Everyone Prices In

Take rates are how marketplaces charge for access. Amazon charges 8-15% referral fees across most categories, while Etsy charges 6.5%. Shopify's transaction fee on its basic plan is intentionally minimal. These figures are outlined in contracts, detailed in help documentation, and included in every seller's initial cost model.

The rationale for accepting these fees has always been straightforward: platforms control distribution. A brand without an existing customer base pays the platform for access to hundreds of millions of buyers. The fee represents the rent for that relationship.

For the initial years of marketplace growth, this framework held true. Sellers competed on product quality, listing optimization, and price. Organic search within the platform was still effective, and a well-structured listing could reach page one based on merit. The take rate was the complete cost of doing business on the platform.

Around 2018, the algorithm's relationship with advertising shifted, evolving from optional amplification to structural dependency, permanently altering unit economics.

The Second Layer: The Fee That Grew Up Around It

Amazon launched its advertising platform in 2012. Initially, sponsored product listings served as an optional accelerant: a way to quickly boost visibility for new products or drive volume during peak seasons. For the first few years, this description was accurate.

As more sellers joined the platform, securing organic positions became more challenging. While Amazon's algorithm had always prioritized conversion rate and listing relevance, sponsored placements steadily consumed page-one real estate as category competition intensified. By the mid-2020s, the top four positions in most competitive Amazon searches were sponsored results, blurring the line between "organic" and "paid" in the buyer's experience.

Sellers who stopped advertising saw their organic rankings decline, due to both algorithmic weighting and the increasing volume of competing listings actively paying for placement. The ad-free option became a reliable route to low visibility.

The extent of this shift is now quantifiable. Amazon's advertising revenue reached $56.2 billion in 2024, an 18% year-on-year increase, and grew to $68.6 billion in 2025. Advertising accounted for 9.36% of Amazon's total revenue in Q2 2025, its highest proportion on record. The average cost-per-click for sponsored products stood at approximately $1.04 in 2025, continuing its upward trend. Sponsored product ad prices were 48% higher in Q1 2025 than in Q1 2019.

Currently, more than 70% of Amazon sellers run active ad campaigns, up from 40% five years ago, illustrating the erosion of optionality.

The result of this two-layer accumulation is that sellers in competitive Amazon categories typically achieved net margins of 5-15% by 2023, with highly competitive categories tending toward the lower end. In contrast, Shopify merchants operating direct-to-consumer, without a marketplace take rate or mandatory advertising dependency, typically achieved 10-20% in the same period. The gap is almost entirely explained by the platform's cost structure, rather than fundamental differences in operating efficiency.

Total fees paid to Amazon by sellers exceeded $150 billion in 2024. Amazon's retail business is now just 40% of the platform's transaction volume. Fees, advertising, and logistics services are the core business. On April 2, 2026, Amazon announced a 3.5% fuel and logistics surcharge on FBA fulfilment fees, effective April 17. The structural direction has not changed.

The Pattern Holds Across Platforms

Amazon's trajectory is the most visible, but the architecture is common across the marketplace category.

Google's shift toward paid placements in organic search followed a similar path over the same period: establish a search product, build distribution, and then introduce paid amplification once the distribution becomes too valuable for advertisers to forgo. App store economics on Apple's iOS App Store carry a standard 30% take rate on digital transactions, a figure Google Play has scaled back from under regulatory pressure. The EU's Digital Markets Act has already compelled Apple to allow alternative stores in Europe, and multiple jurisdictions continue to challenge both platforms' fee structures.

The consistent pattern is this: build the platform, subsidize distribution during the growth phase, and then introduce paid amplification once distribution becomes the only viable path to buyers. Take rate is step one. Advertising is step two. They are different line items on separate invoices, but they contribute to the same structural outcome.

This is what the take rate framing consistently overlooks. When a platform is said to charge 15%, that's the cover charge. The advertising spend is the minimum required payment once you're engaged.

The Third Layer: AI Agents and the End of Platform Discovery

The two-layer model has held because platforms controlled discovery. Shoppers searched within Amazon, scrolled within TikTok Shop, and browsed within Etsy. The platform controlled the moment a buyer decided what to look for next, and advertising spend bought visibility within that specific context.

AI-powered shopping agents are now contesting that control.

Tools like Perplexity's Comet browser and other agentic interfaces operate outside the platform's discovery stack. Rather than a buyer opening Amazon and searching for "wireless earbuds," an agent queries multiple sources, evaluates options based on criteria, and presents a shortlist, potentially bypassing sponsored placements, platform-specific ranking algorithms, and in-marketplace ad spend entirely.

Amazon's response was revealing. In November 2025, the company sued Perplexity, alleging that its Comet browser accessed Amazon accounts and initiated checkout flows without authorization. A court granted Amazon a preliminary injunction in March 2026 and Perplexity filed its appeal with the Ninth Circuit on April 1, with Amazon's response due April 22. Walmart and Shopify have also implemented guidelines prohibiting agents from directing users straight to checkout. Platforms are treating agentic access as a terms-of-service violation, which tells you something about how they assess the structural risk.

McKinsey's analysis of agentic commerce suggests that this model could directly undermine retail media networks built on ad-based revenue, as agent-driven discovery bypasses sponsored placements without the buyer ever seeing them. eMarketer projects that AI shopping platforms will account for 1.5% of total e-commerce sales in 2026, approximately $20.9 billion, nearly four times the 2025 figure. Some brands are already reporting that AI agents account for 10% of their incoming revenue.

For these brands, the advertising spend required to maintain platform visibility yields less reach than before. An agentic buyer doesn't need to engage with the platform's discovery layer at all. If discovery occurs externally, neither the take rate nor the advertising revenue function as intended.

What Operators Should Track Now

The platform fee structure is shifting beneath sellers who haven't yet adjusted their unit economics to account for the full stack. The practical audit involves tracking three key metrics:

• Effective take rate: The percentage of gross revenue paid to the platform in total fees, including all advertising spend. For most Amazon sellers in competitive categories in 2025, this ranges from 40-55%. Note the gap between this figure and the referral fee displayed for your category. A significant and growing gap indicates a mandatory advertising environment, where your product is priced based on only half the cost structure.

• Advertising attribution: The percentage of your platform sales that require paid amplification. A 30-day test with zero ad spend will reveal this. If organic volume doesn't compensate for the loss of paid campaigns, advertising is a structural operating cost and should be treated as a fixed percentage of revenue rather than a discretionary cut when margins tighten.

• Agentic discoverability: Whether your product data appears in AI-powered shopping interfaces, structured comparison tools, and external discovery surfaces. A brand optimized solely for Amazon's internal algorithm but absent from external product graphs is building a distribution dependency on a single discovery layer that is now being challenged from multiple directions.

Companies building durable margins on marketplace infrastructure in 2025 generally pursue two strategies simultaneously: reducing reliance on the platform's discovery layer by cultivating owned audiences and direct purchase channels, and investing in structured product data that extends beyond the platform into external discovery surfaces. The same two-layer dynamic made most PE-backed marketplace rollups so fragile. TIL's private equity coverage documents exactly how that played out.

A better operating framework treats the marketplace as one distribution channel rather than the entire distribution function, incorporating the two-layer cost into unit economics from the outset, rather than discovering it during a margin compression review twelve months later.

Disclaimer

The Industry Lens is an editorial publication. Everything here is analysis and opinion, produced to inform strategic thinking and provoke useful questions. It is not business, legal, financial, or technical advice of any kind. Before making decisions about your distribution strategy, pricing model, or platform economics, do your own research and consult qualified professionals who know your specific situation. We provide the framing. The specifics are yours to work through.