AI & Machine Learning: The Fundamental Guide for Google Search

Yamil Amed Abud

4 years ago

[display-name-category]

[post_author]

AI and machine learning have gotten a lot of buzz this year. You might be wondering if this is relevant to your ads. Yes, it is! Both features are new, revolutionary strategic approaches to managing search campaigns in 2020. 

Below is a fundamental guide that will help you understand how you can leverage each of these features to reach your Google Search campaign goals. 

Understanding Smart Bidding

Before we get into the solutions, let’s freshen up on how automated bidding strategies work on the back-end.

1. The Bidding Process

Let’s assume somebody in Miami is looking for condos in Jackson Hole. Here is the type of data Google considers to make a decision:

For every variable in a keyword auction, Google will assign a probability of conversion. In this case, the user is on a mobile phone in North Miami. According to the advertiser’s past data, this scenario has been successful. 

Sounds set, right? Nope, there’s more. Google also noted that the user is only 18 years old, which makes them less likely to have enough money to buy a Condo in Jackson Hole. Given all the variables, Google determines the user to be of low value and decides to bid only $1. As a result, the bid is not high enough against the other advertisers, and the ad is not triggered. 

When we discuss ‘variables,’ we refer to anything within a campaign segment that influences Google Ads’ decisions. For instance, a section can be “Locations,” and the variables within “Locations” are countries, states, or cities. Another section can be “Devices,” and the variables within “Devices” are computers, tablets, or mobile. Google Search adjusts bids for every keyword auction in real-time, considering all variables in less than one second.

2. AI Efficiency

The fewer variables you have within one campaign, the faster it learns. To understand efficiency, let’s assume that we have three identical systems that can only process ten entries per second. 

Below, you will find each system within a different scenario:

You’ll note that as you increase the number of variables, the longer it takes to review data. The closer you are to the processing limit, the less efficient your system.

3. Data Volume and AI Learning Time

The machine needs to gather enough data for each variable. Once it has sufficient information, it understands which variables drive conversions and which don’t. 

Below are two examples of how AI would perform under different scenarios: 

What this shows is that Scenario 1 can determine which variable is more cost-effective and more likely to convert with fewer impressions. In contrast, Scenario 2 lacks the data to understand which variable best serves the campaign.

4. Interesting Fact about AI

If you have a campaign targeting the United States, Google does not assign a conversion probability to only one variable. Instead, it segments the region you are targeting in States, Cities, and Zip Codes. Below you can find a video shared by the Google Developers, where they explain how AI maps data:

Search Campaigns Solutions

Now that we’ve reviewed how Smart Bidding works, we can better understand the settings. Depending on your objective, you can focus your efforts on either of these strategic solutions:

Solution 1: Reducing Variables

Splitting campaigns per location will reduce the number of variables and improve the efficiency at which Google adjusts the bids. The more you divide up the campaign, the better. That said, you will want to split it up in a way that still has enough data over the past 30 days. 

In the example below, the user launched Florida, California, New York, and Texas as separate campaigns because each had enough conversion data to stand alone. For the states lacking data, the user grouped them on one campaign labeled “Other States.”

 

Solution 2: Driving Traffic

Using match types such as “exact” or “phrase” limits your keyword bidding strategy. Separately, creating Ad groups with “full-broad” keywords that form a “theme” drives more traffic to your campaigns. 

In the example below, we first let AI indicate which keyword match types drive higher conversions and then used the themed ad groups with full-broad keywords based on the match types that work best. 

Solution 3: Adjusting Average CPCs Manually with Target CPA

Now that the campaigns are fully automated, we can set average bids using Target CPA. Considering that CPA and CPC are directly proportional, you may use your campaign’s past data to mirror a specific CPC through your Target CPA.

In the image above, you can see how CPA and CPC behave similarly. Therefore, if you wanted to mirror a CPC of $4.02, you would need to set a Target CPA of $22.90. 

Conclusion

Although many believe  that automation is surrendering control of your campaigns, this isn’t true. Understanding how artificial intelligence works is essential to leveraging it for your campaigns.

To learn more about how you can implement AI to your digital efforts, reach out to one of our strategists today.