What is Google FeedGen
FeedGen uses state-of-the-art LLMs to improve product titles, generate more comprehensive descriptions and fill missing attributes in product feeds. This increases the quality of the feeds and helps detect and resolve quality issues. Better product titles provide more visibility in Shopping results, a higher CTR and higher conversion rate. As a result, you can significantly improve revenue without sacrificing your profit. The adjustments can also prevent products from being rejected, thus prevent losing potential sales.
How FeedGen works
FeedGen is a solution that can be connected to your existing Adchieve Shopping feed. Instead of using simple if/then conditions, extensive instructions (prompts) are provided. Adchieve will work with an advertiser's product expert to give some examples to the AI model of how the ideal title should be composed. This creates highly personalized results, ensuring higher quality and consistency. With this, Adchieve can not only solve problems in the feed, but also create completely new attributes that were not in the original input feed, based on the examples provided.
Scoring
Adchieve assigns a score between -1 and 1 for generated titles and descriptions. This scoring system indicates quality and helps you review and validate content before it goes live. After all, you want to avoid adding values to your titles or descriptions that are not accurate. Titles or descriptions where a value has been hallucinated (i.e. generate inaccurate or irrelevant content) are given a score of -1. When words have been removed it gets a score of -0.5, this is because you want the titles to contain more information and not less. When there are no changes, the result gets a score of 0. Has the title been optimized? Then a value of 0.5 is assigned. If an attribute column is then also added, such as style, then it is assigned a value of 1.
Adchieve A/B test tool
Tip: To ensure that the new titles and descriptions perform better than the old ones, the data is not added as a supplemental feed. Instead, the GMC A/B test tool is used to run a pure A/B test. This tool automatically changes the titles from the original variant to the improved variant and adds a custom label value. Original titles get the value "A," improved titles get the value "B." This allows you to easily make comparisons in your Google Ads reports.
Results
For a client in the electronics segment, we set up a test using the aforementioned A/B testing feature in Adchieve. The test produced the following results in 90 days for Variant A (Original) and Variant B (AI optimized):
● Increased number of clicks: Variant B saw a 9.52% increase in clicks compared to Variant A, indicating a more effective trigger for user interaction.
● Increased number of impressions: There was a 12.55% increase in impressions for the AI-optimized feed, indicating better visibility in search results.
● Slightly decreased CTR: The click-through rate (CTR) decreased by 2.52%, which could indicate a slightly decreased relevance of the ads to the target audience, despite the increase in overall visibility.
● Lower cost per conversion: The cost per conversion decreased significantly by 20.34%, indicating that the AI-optimized feed was more efficient in generating conversions at a lower cost.
● Increased conversion rate: There was a 13.95% increase in conversion rate for variant B, indicating that a higher percentage of clicks resulted in conversions.
● Lower average cost per click (CPC): The average CPC decreased by 11.43%, indicating that the AI-optimized feed was more cost-effective in generating clicks.
In summary, although CTR decreased slightly, the AI-optimized feed (variant B) resulted in more clicks, impressions, a higher conversion rate, and lower cost per conversion and per click, indicating that FeedGen's optimization contributed to a more efficient and cost-effective campaign.
| Variant | Clicks | Impr. | CTR | Avg. CPC | Cost/con | Conv. rate |
| variant a (original) | 2,699 | 226,395 | 1,19% | 0,7 | 19,42 | 3,37% |
| variant b (AI optimized) | 2,956 | 254,815 | 1,16% | 0,62 | 15,47 | 3,84% |
| difference | 9,52% | 12,55% | -2.52% | -11,43% | -20,34% | 13,95% |
Conclusion
The use of FeedGen in practice clearly demonstrates that AI-optimized feeds have the potential to make a significant positive impact on PPC campaigns. With improved visibility, more efficient click rates and higher conversion rates, this technology offers significant benefits to advertisers using Google Shopping.
However, it is important to remain alert to its challenges, such as the need for manual review and fine-tuning of feeds. All in all, the integration of AI into PPC strategies promises an exciting future for digital marketing.