• Skip to primary navigation
  • Skip to main content
  • Skip to footer

Seng Nickerson | Digital Ads Strategy & Growth

Helping businesses spend smarter and grow faster with clear, actionable advertising strategies.

  • Home
  • About Seng
  • Blog
    • Shopping Ads & Product Feeds
    • Search & Text Ads
    • Strategy, SEO & Growth
    • Social & Display Ads
    • Ad Tech & Automation
    • Conversational & AI Ads
  • Resources
  • Contact
  • Free Guide
You are here: Home / Shopping Ads & Product Feeds / How to Improve Product Feed Quality and A/B Test Shopping Ads with Google’s New Experiment Platform

How to Improve Product Feed Quality and A/B Test Shopping Ads with Google’s New Experiment Platform

December 4, 2025 by Seng

Product feed quality is the foundation of every successful Shopping ads strategy, regardless of platform. Google Shopping, Microsoft Advertising, Pinterest, and other comparison shopping engines rely entirely on structured product data to determine when, where, and how your items appear to shoppers. Unlike keyword-based text ads, Shopping ads do not use a traditional “quality score” tied to bidding values. Instead, the strength and clarity of your product feed is your score. Weak, incomplete, or inconsistent product data leads to suppressed impressions, lower click-through rates, poor conversion performance, and disapprovals—issues that can directly impact revenue.

Table of Contents

Toggle
  • Introducing Google’s New Product Data Experiment Platform (PDEP)
  • Why Feed Health Is the Single Most Important Driver of Shopping Ad Performance
    • 1. Clarity and completeness of product attributes
    • 2. Accuracy and compliance
    • 3. User experience signals based on feed content
  • What You Can Test Now With Google’s A/B Product Data Experiments
    • A/B Testing Product Titles
    • A/B Testing Product Images
    • What You Can’t Test Yet
  • Using Google’s Shoptimizer to Automatically Improve Feed Quality
  • Best Practices for Improving Product Feed Quality Before Running Experiments
    • Submit strong baseline content
    • Only test one attribute at a time
    • Prioritize clarity over keyword stuffing
    • Allow tests to run their full duration
    • Avoid making feed changes mid-experiment
    • Watch different stages of the funnel
  • Example Ideas for Experiments to Run
  • The Future of Shopping Optimization Is Data-Driven and Feed-First
  • Smarter Ads. Less Overhead.
    • Weekly Newsletter


For years, I’ve seen firsthand how merchants struggled without a structured way to test title or image variations. In my enterprise work building large-scale feed systems, I’ve consistently championed the need for controlled experimentation to improve product data quality. Seeing Google release a standardized A/B testing platform marks an important step forward for the entire industry. Google, Bing, and Pinterest provide best practices, but none have historically offered a way for merchants to systematically test feed variations and see what actually improves performance. Today, that changes. Google has introduced a groundbreaking capability inside Merchant Center: A/B testing for product titles and product images. This is the first time merchants can scientifically measure how changes to product data influence Shopping Ads outcomes while also in turn impacting Free Product Listings.

This post explores why feed health is critical, how Google’s new Product Data Experiment Platform works, what types of improvements you should consider testing, and how tools like Google’s experimental Shoptimizer can help automate better content. If you run Shopping ads or manage product feeds, this is the moment to level up your competitive advantage.

Introducing Google’s New Product Data Experiment Platform (PDEP)

Google Merchant Center’s Product Data Experiment Platform (PDEP) is the industry’s first native tool that allows merchants to run statistically sound A/B tests on product data attributes. Instead of relying on intuition or historical platform feedback, businesses can now validate exactly which title structures, image styles, or content patterns perform best. Text Ads had it first with Drafts & Experiments, but all of this was housed on the Google Ads side.

PDEP splits traffic using a scientific user-based (cookie-based) method. Half of users see your original data (the control), while the other half see the new version (the treatment). Because this split happens at the user level—not at the product level—the results are more sensitive and meaningful, even when the number of tested SKUs is relatively small. And you don’t even need to mess with your Google Ads campaign settings.

Google’s first two supported experiment types are:

  1. Title Experiments: Test alternate product titles
  2. Image Experiments: Test alternate primary images

This enables powerful new testing patterns, such as comparing a standard white-background image against a lifestyle photo with a model, or comparing long attribute-rich titles against short scannable ones. For brands selling apparel, beauty products, accessories, home goods, or anything visually driven, this opens doors to measurable performance improvements that were previously impossible to isolate.

Even more interesting: for image testing, Google resets past performance bias by creating an invisible “adjusted” version of your existing image. This ensures your new test image isn’t penalized simply because it’s newer and hasn’t yet accumulated historical performance. Both images start from a level baseline, which means the results reflect customer preference—not algorithmic relics.

Why Feed Health Is the Single Most Important Driver of Shopping Ad Performance

Shopping algorithms evaluate thousands of signals to determine which products appear in which auctions, but nearly all those signals depend on the accuracy and completeness of the product feed. The platforms do not modify or “fix” your product content. If your data is vague, inconsistent, incomplete, or misleading, you lose visibility.

Across Google, Microsoft, and Pinterest, the three content pillars remain consistent:

1. Clarity and completeness of product attributes

Shoppers want fast answers—size, color, material, brand, category, gender, style, and use case. Algorithms prioritize content that reduces ambiguity.

2. Accuracy and compliance

Title spam, unnecessary keywords, promotional text, incorrect GTINs, invalid MPNs, and inconsistent category taxonomies can cause disapprovals or lower rank.

3. User experience signals based on feed content

CTR, conversion rate, and competitiveness all flow from product-level attributes. Cleaner content leads to cleaner matching, which often leads to lower CPCs and higher ROAS.

Shopping platforms cannot fix your titles. They cannot enrich your descriptions. They cannot rewrite your product type taxonomy. They evaluate what you provide and determine whether your data deserves reach. High-quality feeds consistently outperform those that rely on defaults.

This is why structured testing—like PDEP now enables—is so critical. You no longer have to rely solely on platform guidelines or guesswork. You can actually experiment with data variations and measure their impact.

What You Can Test Now With Google’s A/B Product Data Experiments

Google’s initial rollout focuses on two high-impact attributes: titles and images.

A/B Testing Product Titles

Testing different title styles can reveal meaningful differences in CTR, conversion rate, or conversion value. For example, merchants may test:

  • Short scannable titles vs. attribute-dense titles
  • “Brand + Product Type” vs. “Product Type + Attributes + Brand”
  • Adding missing key attributes like model, material, or use case
  • Moving the most important attribute to the front
  • Removing unnecessary or redundant descriptors

If your catalog contains apparel, furniture, electronics, or multi-variant SKUs, even small title refinements can improve matching quality.

A/B Testing Product Images

Images drive emotional engagement and click-through behavior. Google’s new experiment type lets you test:

  • White background vs. lifestyle image
  • Model vs. no model
  • Close-up detail shot vs. full product shot
  • New angles or new lighting setups
  • Images without overlays vs. images that previously triggered warnings

Because Google neutralizes historical bias, the results reflect real customer preference.

What You Can’t Test Yet

PDEP does not yet support testing:

  • Descriptions
  • Product types
  • Pricing
  • Promotions
  • Multiple variables in the same experiment

However, based on the structure, Google is clearly setting up a future that allows more granular testing across additional attributes.

Using Google’s Shoptimizer to Automatically Improve Feed Quality

Alongside the experimentation platform, Google has also shared an optional open-source tool on GitHub called Shoptimizer. While it is unsupported and comes with standard use-at-your-own-risk disclaimers, it provides an important signal: Google recognizes that merchants need better tools to optimize feed content before it enters Merchant Center.

While it doesn’t require a full engineering background, you will need someone technical enough to understand and work with APIs. Shoptimizer serves as a REST API that takes product data (in the same format as a Content API for Shopping request) and returns optimized data. It includes built-in optimizers that:

  • Enforce attribute length limits
  • Remove promotional text
  • Mine missing attributes (brand, color, gender, size)
  • Improve descriptions
  • Clean up invalid characters
  • Swap out poor-quality images
  • Remove invalid MPNs
  • Fix misrepresented conditions
  • Ensure product types are compliant
  • Detect and exclude prohibited shopping items

For merchants with engineering resources, Shoptimizer can be integrated into a pipeline to auto-sanitize and optimize data before it reaches Merchant Center. For merchants without engineering support, Shoptimizer still serves as a valuable reference. You can review the logic and incorporate the concepts manually:

  • Ensure titles are clear, within length limits, and include key mined attributes
  • Review image quality and remove overlays
  • Fix identifiers and clean up data inconsistencies
  • Expand descriptions using key attribute data
  • Avoid excessive or irrelevant attribute lists

Even without adopting the API, understanding Google’s optimization logic helps you produce stronger product data.

Best Practices for Improving Product Feed Quality Before Running Experiments

Before testing variations, it is essential to ensure your baseline feed is healthy. A/B tests can only provide value if the foundation is clean.

Submit strong baseline content

Start with accurate identifiers (GTINs, brand, MPN), high-quality images, and clear titles. Fixing the basics should come before experimentation.

Only test one attribute at a time

If you test titles and images simultaneously, you cannot isolate the cause of the performance change.

Prioritize clarity over keyword stuffing

Shopping does not reward keyword stuffing. Clear, structured titles win.

Allow tests to run their full duration

Google recommends 4–6 weeks to smooth out weekly shopper behavior patterns.

Avoid making feed changes mid-experiment

Data contamination resets learning windows and ruins statistical significance.

Watch different stages of the funnel

A higher CTR doesn’t always lead to higher ROAS. A more descriptive title may reduce clicks while increasing conversion rate. Always evaluate tests based on your core business goal.

Example Ideas for Experiments to Run

Here are several practical experiment ideas merchants can begin testing immediately:

  • Compare white-background images vs. lifestyle images with a model.
  • Test moving size or color to the front of the title for apparel.
  • Test shortening long titles to reduce truncation on mobile placements.
  • Test descriptive titles for home goods vs. minimal titles.
  • Test different main images for products with multiple angles or textures.
  • Test adding key attributes to descriptions manually following Shoptimizer principles.

Each test delivers actionable insight you can apply across your entire catalog—amplifying the ROI of even a single experiment.

The Future of Shopping Optimization Is Data-Driven and Feed-First

For the first time, Google is enabling merchants to scientifically validate changes to their product feed. This is a major shift in how Shopping ad optimization will be done moving forward. Businesses that adopt clean data practices, structured testing, and attribute-level optimization will outperform those relying on static, untested feed structures.

Feed health isn’t glamorous—but it is one of the most powerful growth levers for ecommerce performance. With Google’s new experimentation platform and tools like Shoptimizer, merchants now have the ability to refine content, improve customer engagement, and maximize return on ad spend with far more precision than ever before.

Smarter Ads. Less Overhead.

Stay ahead of the curve with curated insights on digital advertising, AI, and automation. Each update is designed to help you:

  • Reuse proven strategies the big brands pay agencies for
  • Leverage AI tools to simplify marketing (without adding headcount)
  • Avoid the most common (and costly) mistakes when scaling ads

Get practical, actionable updates—delivered straight to your inbox.

Weekly Newsletter

No spam, ever. Unsubscribe anytime.
View our privacy policy.

Filed Under: Shopping Ads & Product Feeds, Strategy, SEO & Growth

Footer

SengNickerson.com

  • About
  • Advisorship
  • Resources
  • Seng on YouTube

Projects

  • Home & Style Blog
  • Cooking with Lane
  • Elusive Pursuits
  • Chasing Experiences Vlog

Legal

  • DMCA Policy
  • Cookie Policy
  • Terms of Use