Artificial intelligence in marketing is no longer the music of the future but daily reality. Plenty of companies jump on the Performance Max or Meta Advantage+ wave, but the results don’t come. Are you wondering why competitors’ campaigns generate orders while yours only spend the budget? The answer almost always lies in the quality of the data you “feed” the algorithms.
Summary for Those Who Don’t Have Time to Read the Whole Article
- Both Performance Max and Meta Advantage+ need hundreds of data signals to function, not just keywords.
- Meta’s new Andromeda algorithm changes the rules. You no longer handle targeting manually, but through the quality and variety of your creatives.
- Without server-side measurement (CAPI) and enhanced conversions you lose tens of percent of data, which significantly worsens the learning of the algorithms.
- Data volume is decisive. Campaigns with fewer than 30 conversions a month struggle with stable performance.
- The marketer’s role has changed. They no longer set the CPC, but take care of input data quality and strategy.
Does it sound too complicated? We’ll be happy to help.
Google Performance Max: AI Needs Clear Signals
Performance Max (PMax) campaigns automate everything from bidding to channel selection (YouTube, Search, Gmail). But for the system to know who to show the ad to, it needs quality inputs.
Audience Signals Replace the Old Targeting
You used to tell Google: “We want to target people searching for cheap shoes”. Today you have to supply it with Audience Signals. These are your own data – lists of real customers, website visitors or segments of people with high purchase intent. The system uses these signals as a springboard for finding new, similar users.
Why Doesn’t a Campaign Work Without Conversions?
The algorithm learns by repeating success. If a campaign generates little data, the AI has nothing to draw on.
- For basic operation a campaign needs at least 30 conversions a month.
- For optimal performance and a stable return on ad spend (ROAS), the ideal threshold is around 150 conversions a month.
Important: it is not enough to measure only the “number” of orders. You must also send the conversion value to the system. That way the AI learns to prioritise customers who make a big purchase over those who buy only a small item.
You might also be interested in: Measurement Trends and a Tracking Plan for 2026
Meta Andromeda and Advantage+: Creative as the New Targeting
On Facebook and Instagram there has been a fundamental change in how the system selects ads (the so-called retrieval engine). The new Andromeda system uses powerful artificial neural networks and changes the way Advantage+ campaigns are worked with.
Data and Creative as the Glue of the System
With the arrival of Andromeda, a complex campaign structure and detailed interest targeting stop being important. What is decisive is the variety of creatives.
- Hierarchical indexing – the system can analyse the content of your videos and images. If there is a specific product in the video, the algorithm itself finds the people who are interested in that product, better than you could with manual targeting.
- Impact on results – advertisers using Advantage+ creative supported by Andromeda can see an increase in return on ad spend (ROAS) of up to 22%.
For this to work, you must supply the system with 5–10 different creatives (video, static, various formats). The system needs material so that it can test what works on whom.
Server-Side Tracking and Advanced Matching
This is the part that many companies underestimate, yet it is absolutely fundamental for AI to work. Because of privacy protection and ad blockers, the ordinary measurement code (pixel) in the browser is no longer enough. It sees only part of reality.
1. Meta Conversions API (CAPI) and Advanced Matching
For Facebook (Meta), deploying server-side tracking (CAPI) is a necessity today. Order data is sent straight from your e-shop’s server to Facebook’s server. This way you bypass outages in browsers.
At the same time it is necessary to turn on Advanced Matching. Along with the conversion you send encrypted customer data (hashed email, phone). Meta links this data with its user database and assigns the conversion to a specific person more accurately. You thereby raise the quality of the data the AI learns from. It is of course better to set this function up manually, so as to achieve high relevance and data quality.
2. Google Enhanced Conversions
The same principle applies at Google. The Enhanced Conversions feature makes it possible to securely send customer data back to Google Ads. This helps the system link a conversion to an ad click even in cases where cookies fail. Diagnostics in Google Ads report on average an increase in conversions in the higher single-digit percentages.
You might be interested in: Cookie tips from the analyst’s practice
3. Sklik and Data Accuracy
Even the Czech Sklik requires precise data for its automatic strategies and retargeting. Without a correctly deployed conversion code that records the real values of orders, the system cannot effectively deliver ads to the right people in the content network. Sending users’ personal data to increase the accuracy and quality of the data is a given, as are Advanced Matching and Enhanced Conversions.
Don’t be afraid to leave setting up advanced measurement to us.
Conclusion
Today it is not the one who spends hours tweaking campaign settings who wins, but the one who can feed the algorithms the best data. Without accurate measurement and without quality signals, the AI in your campaigns will only be guessing.
But if you take care of the technical side and back it up with top-notch creative, automation can bring you results you would never reach with manual management.
Frequently Asked Questions
Is a small budget enough for automation to work?
It is not so much about the size of the budget as the number of orders. For the artificial intelligence to learn what works, it needs to see at least 30–50 purchases a month. If you have fewer, the system will just guess and the results will fluctuate.
Does it matter if I don’t have advanced measurement (CAPI) set up?
Yes, a lot. Without it you won’t see purchases from people with iPhones or those who use ad blockers. The advertising system will then think your ad doesn’t work and stop showing it, even though it is actually earning you money.
Why does Meta Advantage+ ignore my interest settings?
Meta’s modern algorithms (powered by the Andromeda system) treat the interests you enter (e.g. “fitness”) more as a loose recommendation. If, based on data, the system finds that the conversion is likely to be made by someone outside this group, it will prioritise the result over your settings.
Does this mean I no longer have to deal with targeting?
Manual targeting is receding into the background. Your main targeting tool becomes the creative (the content of the ad) and the signals (customer data) that you send to the system. The right creative attracts the right people.
Sources
Search Engine Journal (2025) ‘Google Performance Max: Everything You Need To Know’, Search Engine Journal, [online]. Available at: https://www.searchenginejournal.com/google-performance-max/498283/
Google (2025) ‘Enhanced conversions for web impact results’, Google Business Help Center, [online]. Available at: https://business.google.com/us/accelerate/resources/articles/enhanced-conversions-for-web-impact-results/
Google (2025) ‘Set up enhanced conversions for web using Google Tag Manager’, Google Ads Help Center, [online]. Available at: https://business.google.com/us/accelerate/resources/articles/set-up-enhanced-conversions-for-web-using-google-tag-manager/
Google (2025) ‘Configure the Google tag for enhanced conversions for leads’, Google Ads Help Center, [online]. Available at: https://business.google.com/us/accelerate/resources/articles/enhanced-conversions-leads/
Engineering at Meta (2024) ‘Meta Andromeda: Supercharging Advantage+ automation with the next-gen personalized ads retrieval engine’, Meta Engineering Blog, [online], 2 December. Available at: https://engineering.fb.com/2024/12/02/production-engineering/meta-andromeda-advantage-automation-next-gen-personalized-ads-retrieval-engine/
MediaPost (2024) ‘Meta Discloses “Andromeda” Ad Retrieval’, MediaPost, 3 December. Available at: https://www.mediapost.com/publications/article/401591/meta-andromeda-ad-retrieval.html