Audience Label

Introduction

The Audience Labels model evaluates the performance of each audience segment in Amazon DSP delivery across critical dimensions such as reach, purchase volume, Detail Page Views (DPV), and conversion efficiency. By analyzing these key metrics, advertisers can identify high-performing segments, uncover untapped potential audiences, and optimize their targeting combinations to capture high-value shoppers.

Note: This model is only applicable to AMC Instances that include active DSP advertisers.

Questions this model can answer:

Untapped Opportunities: In DSP ads, which untargeted label audiences are worth testing for delivery?
Audience Profiling: What are the portrait characteristics of the audiences reached by current DSP ads?
Segment Performance: How do different characteristic labels (Demographic, Lifestyle, In-Market) perform relative to one another?

When should you use it?

DSP Strategy Optimization: Use this model to judge the performance of existing audience labels and discover more potential labels to cover valuable audiences.
Brand-Level Insights: Gain deeper understanding of user characteristics to refine your brand's overall shopper persona.
Campaign Optimization: Provide actionable guidance for DSP account managers by identifying specific segments for budget reallocation or bid adjustment.
ASIN-Specific Audience Discovery: Use custom models to track specific ASINs. This allows you to see how audience performance varies between different product lines in your portfolio.

Best practices using Compose Model Query

Scenario	How to set up	Metrics to include	What to look for	Strategic Decision
Untapped Segment Discovery	Filter by Target Status: Not Targeted and define an Audience Sizemagnitude.	Total ROAS, Reach UV, eCPC UV.	Labels in the lower-left corner of the bubble chart (dotted bubbles).	Add these "high-potential" segments to a test campaign for incremental growth.
Product Line Differentiation	Use Customized ASINsto focus on a specific product category.	Total DPV UV, Total Purchase UV.	Performance differences between labels for Product A vs. Product B.	Establish distinct targeting personas for each major product line.
Persona Deep Dive	Filter by Audience Type: Demographic or Lifestyle.	Reach UV, Total Cost, Impressions.	General brand-level insight and coverage scale.	Use these for creative direction and storytelling, even if ROAS is lower than niche tags.

Data Logic and Model Limitations

Understanding the thresholds and logic of this model is essential for accurate strategic moves:

Magnitudes and Comparability: Comparisons of label performance are only referential when audience magnitudes are similar. Large differences in coverage make effect data like clicks and conversions non-comparable.
Data Reliability & Reach: If a label's Reach UV is low, conversion data (ROAS/PR UV) may fluctuate wildly. In these cases, rely on CTR UV and eCPC UV to judge interest until the sample size grows.
Expansion Quality Risk: High-performing labels with low coverage may lose quality when expanded for new delivery attempts. Small-scale trial delivery is recommended before significant scaling.
Privacy Thresholds: AMC automatically replaces sensitive fields with NULL or filters rows if a segment reaches fewer than 1,001 unique users per month.
Historical Data Cap: AMC only stores data for the past year; custom models are limited to a 365-day maximum time range.
NTB and ASIN Specificity: Conversion and New-to-Brand (NTB) metrics are most accurate when tracked ASINs are defined in a custom model. Predefined models reflect overall store performance.

Unique Reach

Path to Conversion

Last modified: 2026-02-23Powered by

Outline