Skip to content Skip to footer

Advertising in Google Ads using artificial intelligence

In this article, I would like to metaphorically demonstrate, using the example of an online store, how Google algorithms are trained using intelligent strategies in contextual advertising, such as Maximize Conversions. Let’s imagine that products in a certain hardware store are internet users who enter the keyword “Home Goods” into the search, and sales represent conversions.

So, what do we have? In the online catalog, all product categories fit under this search query, namely 12,000 products. In our analogy, this is the available audience of various types. At the initial stage, when you just launch an advertising campaign in Google Ads, the system initiates preliminary training over 5 days.

During this period, your ads will be shown to different product categories, as if blindly probing the ground to understand which categories of products will be bought. And then, suddenly, a purchase of laminate occurs, followed by linoleum and parquet. After 5 days of training through Google Ads, let’s say there are 10 sales. Out of these, 7 are laminate, 2 are linoleum, and 1 is parquet.

After the training, the search engine will have an idea of which particular type of product is the most sold on your site, and will focus specifically on this segment in a percentage ratio (laminate 70%, linoleum 20%, parquet 10%). In our analogy, this would be a certain type of audience with specific properties.

  • Gender: Men (80%), Women (20%)
  • Age: 32 – 45 years old
  • Interests: Gardening, Home Improvement, Sports, Cars
  • Active Online Time: 7:45 PM to 10:15 PM
  • Occupation: Mechanic, Hairdresser, Repair Services
  • Places Visited: Boston, Medford, Waterdown, etc.

This is an approximate set of parameters that Google may have at its disposal. In reality, there can be hundreds or even thousands of different characteristics about each user on the network. As the owner of a smartphone with an Android operating system, Google even has the ability to look into your phone contacts. Thereby determining your social circle and creating your digital profile.

As a result, Google will redistribute the main budget in favor of showing your ads for the laminate category. Sales occur more intensely, but specifically certain models of laminate are purchased, mainly 90% from the German manufacturer Tarkett, and the Russian manufacturer Kronostar. The Google Ads algorithm will then filter the laminate by these manufacturers, significantly narrowing the audience reach, as being more inclined to purchase your product specifically. I really hope this analogy is understandable 😅.

So, moving forward… Three weeks pass, sales are increasing, and the cost per conversion is decreasing due to the redistribution of the budget towards more potential customers. Having collected more statistics, Google tries to identify what the products that are being purchased the most have in common.

  • Color: Brown
  • Panel Thickness: 8 mm
  • Surface: Glossy
  • Pattern: Wood
  • Embossing: 4V bevels

Then, the algorithm again refines the audience based on more specific parameters and begins to target the advertising more precisely.

As you’ve understood, intelligent strategies for bid setting and target audience selection are trained precisely according to this principle, constantly delving deeper into the filter to more accurately find potential customers for your internet resource. The more conversions we provide to the advertising system, the more finely the search engine will learn to distinguish good visitors from bad ones (clickers, people not interested in purchasing).

Also, in the advertising account’s general summary, Google may provide some excerpts from the training.

And finally, some advice on choosing events on the site as conversions. I recommend using goals where the visitor shows their contact details or is compelled to write a message:

  • Contact form
  • Phone call (call-tracking)
  • Callback request
  • Online chat

I strongly advise against using micro-goals, such as clicks on elements on the site: phone number, email, messengers, adding a product to the “shopping cart“. Such actions are massively performed by spam bots that come to your site, imitate real users, and carry out fake conversions. The percentage of such events can reach up to 90%. Now, apply these conversions as a filter for products in the catalog, according to the analogy described above, and you will understand how much you dilute the focus of the target audience sample.

Traditionally, at the end, I offer to familiarize yourself with an excellent WhatsApp widget for installation on the site, which solves the problem of fake conversions. What can the widget do?

  • Protection against fake conversions
  • Automatic sending of events to GA4
  • The ability to send offline conversions to Google Ads
  • Notifications about messages in Telegram
  • Log of first messages