Feed management in PPC, especially within platforms like Google Ads and Microsoft Ads, is crucial for organizing and updating product or service data. It serves as the foundation of Shopping Ads, ensuring that viewers are presented with relevant results and accurate information. 

With AI tools like ChatGPT and Google’s FeedGen, you can enhance your product or service data to deliver more relevant and accurate information to your audience. 

This article will explore the core principles of product feed optimization and the capabilities of AI-driven tools, including the benefits and limitations compared to traditional feed management tools.

What is product feed optimization?

At its core, product feed optimization in PPC is about creating keyword-rich titles and descriptions that:

• Match search queries.
• Show information.
• Categorize products properly.
• Add additional information like color, size, etc. 

In more advanced optimizations, you also consider optimizing images, adjusting prices, using performance labels and conducting A/B tests. 

While the best practices can vary depending on the product category, a common approach for creating effective product titles is with the following structure:

Apparel

• Brand + Gender + Product Type + Attributes (color, size, material)

Consumable

• Brand + Product Type + Attributes (weight, count)

Hard goods

• Brand + Product + Attributes (size, weight, quantity)

Electronics

• Brand + Attribute + Product Type

Books

• Title + Type + Format (hardcover, ebook) + Author

To effectively optimize product titles for online retail, here are some best practices: 

• Adopt a customer-centric approach.
  • Think from the perspective of your target audience and understand their shopping behavior.
  • Include key attributes in titles (i.e., color, size, material, and specifications) as customers often use them to refine their searches. 
• Adhere to platform-specific guidelines, especially Google’s strict rules regarding using capitals and exclamation marks. 
• Use data from Google Ads search terms reports, Google Trends, and tools like Semrush or Ahrefs to identify high-performing queries and relevant keywords. 
• Tailor your titles based on the product vertical, considering how customers in that industry typically search. 
• Include brand names in titles if they are well-recognized and relevant, but exclude lesser-known brands that might not resonate with shoppers.

Optimizing a product feed with ChatGPT

To begin with AI-based feed optimization, starti with ChatGPT. 

You can input your feed and optimize the title and description, gradually adding more information. 

This approach provides a foundational understanding of feed optimization’s capabilities and allows you to focus on significant improvements.

You can use popular plugins like ChatGPT in Google Sheets and Docs, or you can use App Scripts to let the ChatGPT API work on that sheet.

A sample query could be: 

• “Write a Google Shopping optimized product title with a maximum of 150 characters using the [title], [description], [category], [color], [size] and [gender].” 

The input feed:

The result:

While this is not an all-time peak performance of product feed optimization, it’s an improvement of the input feed and a good base to work with.

Setting up ChatGPT + Google Sheet is easily done in 15-30 minutes, including a Plugin or AppScripts setup on this basic level.

A more advanced approach can include special input fields to prompts, Google Shopping guidelines and optimization best practices that ChatGPT can use. 

ChatGPT is a cost-effective way to begin feed optimization for quick title and description improvements. However, if you need a more polished and advanced solution, consider FeedGen.

Optimizing a product feed with Google FeedGen

Powered by Google Cloud’s advanced large language models, FeedGen is an open-source tool designed to: 

• Enhance product titles.
• Craft detailed descriptions.
• Supplement missing attributes in product feeds. 

Its main purpose is to aid merchants and advertisers in detecting and rectifying quality shortcomings in their feeds using generative AI presented in an accessible and adjustable manner. 

By leveraging Google Cloud Platform’s Vertex AI API, FeedGen offers both zero-shot and few-shot inference capabilities. 

The few-shot prompting feature, in particular, allows users to incorporate the best samples from their Shopping feed, thus tailoring the model’s output for consistency and elevated quality.

This can be further refined by adjusting the foundational models with proprietary data.

FeedGen presents many advantages, notably its proficiency in swiftly refining product titles and descriptions in large volumes. 

It’s especially beneficial when depending on third-party data and ensures the integrity of titles and descriptions by avoiding fabricated additions. 

Furthermore, users can effortlessly select their preferred titles and descriptions and transform them into a supplemental feed, ready for integration into their Google Merchant Center.

However, using FeedGen can be challenging as it requires the creation of a Google Cloud Project, Vertex API activation, and a solid understanding of Vertex AI for machine learning operations.

The outcome’s quality is contingent upon the quality of your prompts, requiring expertise in defining the optimal prompting examples. Setting up examples for each product category can be time-consuming, and meticulously reviewing all products remains imperative to ensure top-notch quality. 

Currently, only English is supported, although there’s a potential for expansion into other languages through discussions with Google representatives.

Dig deeper: Top 5 data feed errors that can sabotage your ecommerce campaigns

What to expect when working with FeedGen

Marketers use AI for feed optimization to find scalable and valuable solutions. FeedGen fits this need, working well for both small and large feeds. It fills in missing data and produces detailed descriptions. Its setup can be complex, but its potential is vast.

I’ve tested FeedGen and found it reliable. Agencies might even use it for multiple clients. Setup takes under an hour, but choosing the right prompts can take longer, depending on the industry and desired outcomes.

Monitoring FeedGen is essential. I’ve seen errors, ranging from wrong descriptions to titles losing key details. It’s important to ensure good input, as poor data gives poor results. FeedGen isn’t perfect. 

FeedGen has a rating system for products, from -1 to +1, to help review. Still, I recommend checking each product manually. As for costs, FeedGen is affordable. Bigger feeds might cost more, but smaller feeds should stay around or below $100.

From my experience, you can get up to 80% of properly optimized products after finding your ideal prompts and settings, however until I got to 80% it took me a few tries.

That said, if you manually optimize a larger feed or try to cover it based on static optimization rules, the results would take significantly longer or be worse. From this perspective, FeedGen is already a game-changer on scaling feed optimization.

Marketers often use tools like ProductsUp, Channable, DataFeedWatch and more to optimize product feeds. 

With the rise of AI tools like FeedGen, it’s natural to ask: can FeedGen replace these popular tools? 

The straight answer is no. 

While FeedGen is great for enhancing product details, it doesn’t handle the structural tasks that traditional tools do, such as: 

• Linking data.
• Setting up rules for information mapping.
• Working more high-level optimization of titles and descriptions. 

The ideal workflow might be to:

• Use FeedGen to enrich your product feed with details.
• Then, rely on a conventional feed management tool to get everything organized and ready for marketing platforms.

Dig deeper: How to set up feed rules in Google Merchant Center and ensure quality product data

ChatGPT and FeedGen: Your gateway to well-optimized product feeds 

AI-driven tools like ChatGPT and FeedGen can significantly boost your product feed optimization efforts, offering scalability and improved data quality. 

These tools are valuable additions to your PPC toolkit, but it’s important to recognize their specific strengths and limitations. 

Consider how they can work alongside traditional feed management tools to achieve the best results for your advertising campaigns.