The Quiet Shift: What Amazon's Transparency Gap Tells Us About the DSP Market

Understanding the trade-offs between scale and scrutiny in programmatic advertising

Understanding the trade-offs between scale and scrutiny in programmatic advertising

The programmatic advertising market is experiencing a notable realignment. While The Trade Desk commands roughly 20% of the independent DSP market and Google’s DV360 maintains its dominant position, Amazon DSP has quietly emerged as a formidable alternative—particularly for advertisers willing to accept less transparency in exchange for other benefits.

This raises a practical question for marketers: what does it mean when a major DSP doesn’t provide log-level data, and who’s actually comfortable with that trade-off?

The Data Visibility Issue

Log-level data—the granular, impression-by-impression reporting that shows exactly where ads ran, how much was paid, and what resulted—has long been considered table stakes for sophisticated advertisers. This level of detail allows teams to verify that budgets weren’t wasted on low-quality inventory, detect potential fraud, and understand the true path to conversion beyond platform-reported metrics.

Amazon DSP, despite handling an estimated 7.5% of retail media dollars (which translates to billions in annual spend), has notably limited log-level data access compared to competitors. According to multiple ad tech platforms tracking programmatic bidding patterns, some agency holding companies have been shifting meaningful portions of their Q3 spend from The Trade Desk to Amazon DSP—even with this transparency limitation.

The question isn’t whether Amazon lacks transparency. That’s established. The question is why sophisticated advertisers are comfortable with it.

Following the Incentives

The math on Amazon’s fee structure is straightforward. The platform charges no fees for programmatic guaranteed deals on Amazon-owned media and collects just 1% for ads on open web publishers—significantly below The Trade Desk’s roughly 20% take rate. For large advertisers spending millions monthly, this difference compounds quickly.

There’s also the relationship angle. When Omnicom won Amazon’s US marketing business in 2024, industry observers noted that it would likely influence programmatic spending patterns. That prediction appears to have materialized. Multiple sources familiar with programmatic bidding confirmed to AdExchanger that they observed what looked like a double-digit share of expected Q3 spend moving from The Trade Desk to Amazon DSP within Omnicom.

But cost savings and client relationships don’t fully explain the shift. The real differentiator may be Amazon’s first-party retail data—the ability to target based on actual purchase behavior rather than inferred intent. For many advertisers, particularly those in retail and CPG, this closed-loop measurement may matter more than impression-level transparency.

The Sophistication Question

This situation prompts a more interesting consideration: are we redefining what “sophisticated” means in media buying?

Traditionally, sophisticated advertisers demanded full transparency—detailed reporting, independent verification, and control over every aspect of the media supply chain. That model emerged from an era when fraud was rampant and programmatic was the “Wild West” of digital advertising.

But the market has matured. Walled gardens like Facebook and Google have trained advertisers to accept limited visibility in exchange for scale and performance. According to recent data, the DSP market is projected to grow from $38.92 billion in 2025 to $148.92 billion by 2032, with much of that growth coming from platforms that offer performance over transparency.

Some sophisticated advertisers may now be calculating that fraud detection and viewability verification matter less than they did five years ago—especially when advertising on first-party retail properties where the media quality is generally higher. They may be prioritizing performance measurement and cost efficiency over the ability to audit every impression.

The Measurement Challenge

The incrementality testing movement provides context here. Over half of US brand and agency marketers now use incrementality testing to measure campaigns, according to July 2025 data from EMARKETER and TransUnion. Google recently lowered its incrementality testing threshold to $5,000, making this type of causal measurement more accessible.

If advertisers can prove that Amazon DSP drives incremental sales through controlled experiments, does it matter whether they can see every log file? The incrementality test would capture the true lift regardless of the black box nature of the platform.

This represents a philosophical shift from forensic transparency (examining every impression) to outcome-based validation (proving the campaign caused sales that wouldn’t have happened otherwise). The former requires detailed logs; the latter requires good experimental design.

Market Structure Implications

The broader DSP market shows clear concentration. According to recent analysis, three major players—DV360, Amazon DSP, and The Trade Desk—control 86% of market share. While Amazon has the highest advertising revenue, Google’s DV360 maintains the largest market share, suggesting different monetization strategies.

The Trade Desk showed 26% growth in 2024, outpacing both the overall DSP market growth rate (23%) and Amazon’s advertising growth (18%). This suggests The Trade Desk is gaining share in certain segments, even as it potentially loses large accounts like Omnicom.

The market appears to be segmenting by advertiser needs: sophisticated direct-response advertisers who need granular optimization may stick with The Trade Desk, while brand advertisers focused on retail media and closed-loop measurement may migrate toward Amazon.

The Infrastructure Question

There’s another practical consideration: OpenPath, The Trade Desk’s direct publisher connection, bypasses other ad tech intermediaries. This means spending through OpenPath wouldn’t be visible to SSPs and other platforms that typically observe bidstream data.

If Omnicom or other holding companies significantly increased OpenPath usage, it could appear to outside observers that they reduced Trade Desk spending, when in reality they just changed how they bought inventory. The Trade Desk declined to comment on whether Omnicom uses OpenPath extensively, making this impossible to verify.

This highlights a challenge with analyzing programmatic market shifts: much of the data comes from intermediaries who have incomplete visibility. Real spending patterns may differ significantly from what can be observed.

What This Means for Advertisers

For marketers evaluating DSP options in 2025, the Amazon situation surfaces several practical questions:

First, what are you optimizing for? If preventing Made-for-Advertising sites and ensuring brand safety are top priorities, platforms with robust log-level reporting may remain essential. If you’re focused on proving incremental sales lift and are comfortable with Amazon’s first-party inventory quality, transparency may matter less.

Second, how do you measure success? If your organization relies on multi-touch attribution models that require impression-level data, limited transparency is a dealbreaker. If you use incrementality testing or media mix modeling, you can work with less granular data.

Third, what’s the sophistication of your fraud prevention? Amazon’s owned-and-operated properties have inherently less fraud risk than open programmatic exchanges. If most of your spend is on first-party retail inventory, the fraud detection capabilities enabled by log-level data become less critical.

Fourth, what’s the relationship context? The Omnicom-Amazon example suggests that client relationships can drive platform decisions. Holding companies and agencies need to balance getting the best results for current clients with maintaining flexibility for future business.

The Longer View

The DSP market is expected to reach $804.02 billion by 2035, according to recent forecasts. This growth will be driven primarily by retail media networks, connected TV, and other channels where first-party data enables closed-loop measurement.

In that environment, the traditional definition of transparency—seeing every impression—may become less relevant. What matters is proving causality: did the advertising cause outcomes that wouldn’t have happened otherwise?

Amazon’s transparency limitations may seem like a disadvantage in 2025. By 2030, they may be irrelevant if the market fully embraces outcome-based measurement over process-based auditing.

For now, the answer to “who’s comfortable advertising without log-level data” appears to be: more sophisticated advertisers than you might expect, as long as they have other ways to validate performance.