Benchmark Report Methodology and Key Principles

The ANA Programmatic Transparency Benchmark Q1 2025 report is built on a robust, data-driven methodology that leverages impression-level log data (LLD), advanced statistical modeling, and standardized definitions for performance, cost, and quality metrics. This approach ensures transparency, comparability, and actionable insights across the programmatic media buying ecosystem.

Data Collection

The data collection process starts by supply chain vendors (DSPs, Ad Verification and SSPs) providing Log-Level Data (LLD) data feeds which meet specific requirements. These requirements have been defined by TAG TrustNet as part of the guidelines of the TAG Certification For Transparency (CFT). The TAG TrustNet Requirements state which data fields are required or recommended to deliver the ANA Programmatic Transparency Benchmark (the “Industry Benchmark”). 

Each advertiser who participates in the ANA Industry Benchmark has each of their LLD feeds validated against these requirements. There may be, on occasion, specific fields that are not supported by a vendor which can impact the availability of some metrics within the benchmark metrics provided to individual advertisers (the “Advertiser Benchmarks”). 

Participants: 39 marketers enrolled in the Benchmark, with 23 providing complete LLD datasets.

Timeframe: The dataset covers activity from November 1, 2024 to March 31, 2025.

Volume: 41.9 billion impressions and $242 million in ad spend were analyzed.

Access to LLD feeds

To generate the Industry Benchmark LLD is required. LLD provides the most granular level of information about the programmatic auction process and attributes of each individual ad impression including the costs, quality and delivery information included in a single unified transaction record. 

The Benchmark Requirements determine that LLD needs to be generated by each individual vendor on an ongoing basis every 24 hours, in line with the TAG TrustNet Requirements which have been defined in line with industry standards and the EU Digital Markets Act requirements. The LLD can be delivered either directly by the vendor, on behalf of the advertiser, or by the advertiser or agency depending on the legal agreements between the parties.

Data is delivered to the Fiducia LLD Management Platform using several standard methods for transferring large quantities of data, including Amazon S3 buckets, Google Big Query and Cloud Storage (GCP), or SFTP. The data transmission method depends on the requirements and best practices of each vendor. 

​

LLD processing and matching

To produce the required datasets for analysis and benchmarking, LLD feeds from DSPs, ad verification platforms, SSPs and Data Exchange partners need to be matched.

As there is no industry standard for LLD, the formats and field names of each LLD feed differ for each vendor.

To solve this problem, Fiducia maintains a library of custom data connectors for each data provider. This allows the Fiducia LLD Management Platform to ingest, harmonize and match LLD across vendors to produce a single unified record for every ad impression. These unified records are used to come up with the individual Advertiser Benchmarks and for the Industry Benchmark, by aggregating the data of participating advertisers.

There are four types of automated data processing and reconciliation undertaken to generate the Benchmarks.

• DSP only data.

• DSP and ad verification impression matching using unique impression level identifiers available in the DSP LLD and passed to ad verification tags via key value pairs and matched.

• DSP and SSP impression matching using unique ORTB auction impression level identifiers available in the DSP and SSP LLD and matched.

• DSP and 3rd party data provider data matching using publisher site, domain and Seller ID identifiers with the DSP LLD and matched to the 3rd party data provider.

Once data is processed and matched metadata and dimension such as the below are surfaced and stored. The data structuring process allows for analysis, and Benchmarks metric generation.

• Timestamps

• Campaign ID

• Domain/App

• Creative IDs

• Exchange IDs

• Seller ID

• Costs and Fee Data

• Device

• Environment

• Auction Type

Core Benchmark Indices

The Benchmark introduces two core concepts and Indices to help size the evaluate the programmatic ecosystem.

​

TrueImpressions

​A TrueImpression is defined as one that meets all of the following:

• Matched (between DSP and ad verification sources).

• Non-IVT (not invalid traffic or bot activity).

• Measurable (sufficient data available for viewability assessment).

• Viewable according to MRC standards (e.g., 50% of pixels in view for 1+ second).

• Non-Made for Advertising (MFA) is shown in the waterfall using the MFA data provided by Deepsee.io, but is not part of the Benchmark definition of TrueImpressions as it is considered as a subjective value.

​​​

TrueAdSpend Index

• Represents the percentage of total ad spend allocated to TrueImpressions.

• Allows marketers to quantify how much of their budget directly supports quality inventory.
   

TrueCPM Index

• Formula: TrueCPM = Total Spend / Number of TrueImpressions × 1000

• The TrueCPM Index delta measures the gap between standard CPM and TrueCPM.

• This delta reflects the optimization opportunity (i.e., how much spend is going to non-qualifying impressions).​

​

TrueCPM Index simulated delta

• The simulated delta between the CPM and TrueCPM uses a simulation to determine the opportunity marketers can reasonably expect to fulfill by implementing an optimization plan following ANA recommendations and using impression LLD over an always-on platform - such as the Fiducia LLD Management Platform - allowing to manage and monitor closed-loop TrueCPM optimization cycles. 

• To assess the potential for TrueCPM improvement through ad spend optimization all unique combinations of <Campaign ID, SSP, Seller ID, Publisher> are analyzed, and ranked by TrueCPM values. This ranking allows to identify the combinations with the lowest TrueCPM values, which collectively account for approximately 33 percent of total ad spend.

• An hypothetical reallocation scenario is then used to redistribute that 33 percent of ad spend across the top-performing combinations (those representing the ~66 percent of spend with the highest TrueCPM values), in proportion to their current ad spend shares. This allows to estimate the optimized TrueCPM that can be achieved under this scenario. The optimization opportunity is defined as: (Current TrueCPM / Optimized TrueCPM) – 100 percent = TrueCPM Optimization Opportunity.

• This metric indicates the potential uplift achievable through the reallocation of ad spend to TrueImpressions. It is important to note that this model assumes CPMs remain constant post-optimization. In practice, increasing spend in specific channels may lead to higher CPMs, which could impact productivity gains.

• While the TrueCPM Optimization Opportunity does not guarantee specific performance levels, it serves as a valuable directional indicator that supports strategic decision-making.

​

Cost Waterfall

The Cost Waterfall shows the full ad dollar journey based on matched DSP and ad verification data

• Starting at 100% (total ad spend)

• Sequentially deducting:

  • DSP platform costs

  • Data & Other costs

  • SSP platform costs

  • IVT ad spend

  • Non-measurable ads pend

  • Non-viewable ad spend

  • MFA ad spsend

• Ending with TrueAdSpend — the portion that delivers TrueImpressions.

• Key rules:

• Each inefficiency is counted only once.

• Applies average rates rather than medians (used in the detailed findings) to show cumulative impact.

Benchmark Statistical Framework

All detailed benchmark metrics are calculated and graphically represented in the same way. 

Benchmark metrics are analyzed using quartile distributions from all participating marketers

• Top Quartile (best performing)

• Above Median Quartile

• Median (industry midpoint)

• Below Median Quartile

• Bottom Quartile (least performing)

The median is used over averages as averages can be skewed by outliers (e.g., unusually high CPMs). Medians offer a more accurate representation of the “typical” marketer.