What is advertising machine learning? Introducing case studies and operational tips for reducing man-hours and increasing CV by 69%

Automating advertising operations through machine learning can be expected to reduce man-hours and improve advertising effectiveness. However, there are probably many marketing staff who are considering in-house advertising operations using machine learning , but who do not have the knowledge to do so.

In this article, we will explain the outline of machine learning in advertising , the advantages and disadvantages , the necessary period , and the points to keep in mind .

In addition, we will also introduce examples of how machine learning was used to reduce man – hours and increase CV .

Table of contents

  1. What is machine learning in advertising?
  2. What is Machine Learning in Google Ads?
  3. Advantages of using machine learning in advertising operations
  4. Disadvantages of using machine learning for advertising operations
  5. Three points to keep in mind in advertising operations using machine learning
  6. Two successful cases of advertising operations using machine learning
  7. Frequently asked questions and answers about machine learning in advertising
  8. Use machine learning to optimize ad operations


In-house or outsourced advertising?

What is the ideal relationship between advertisers and agencies? We will show you in what cases in-house or outsourcing is appropriate.

What is machine learning in advertising?

Advertising machine learning is AI learning and analyzing huge amounts of data in advertising operations.

It learns various data such as user attributes, keywords searched, and devices , and automatically adjusts bids and selects ads to be delivered based on that data , so advertisers can deliver ads that can be expected to be highly effective with minimal work .

In recent years, the number of companies that use machine learning for advertising operations is increasing due to the desire to “reduce man-hours” and “get more results.”

What is Machine Learning in Google Ads?

In Google Ads , machine learning is primarily used in Smart Bidding and Smart Ads . Let’s take a closer look at each overview.

Smart bidding

Smart Bidding is a function that optimizes the number of CVs and bids for each ad auction . Bids are automatically set according to goals such as “I want to get conversions” and “I want to maximize cost effectiveness” , and ad distribution settings are adjusted.

Bids are optimized based on user signals, so you can expect better results and cost-effectiveness than setting them manually.

Signals are attributes that can identify an individual user, such as location, device used , browser, language, and site activity history.

smart advertising

Smart ads are those in which AI automatically creates and distributes ads based on the budget set by the advertiser Using Google ‘s machine learning technology, creatives suitable for individual users are created and automatically delivered at the optimal timing.

Smart Ads can be divided into the following three types by format.

  • responsive display ads
  • dynamic search ads
  • responsive search ads

● Responsive display ads

Responsive display ads automatically create and serve display ads by uploading assets . Assets are ad headlines, descriptions, logos, images, videos, and more.

The best combination of assets is determined based on predictions that Google derives from past performance For example, if you upload both an image and a video, Google will show the video instead of the image if it thinks the video will perform better.

In addition, since the display format and format are automatically adjusted according to the ad space, it is possible to reduce the time required for ad creation and distribution.

● Dynamic Search Ads

Google Dynamic Search Ads (DSA).gif

Dynamic Search Ads is one of the search advertising functions provided by Google . Google automatically generates ads for users who search for keywords that are highly relevant to the advertiser ‘s web page .

There is no need to set keywords, ad sentences, bids , etc. for each ad .

In addition, since the ad distribution target is determined based on the information in the website , it is possible to display ads to users who could not be reached by keyword-based campaigns .

● Responsive Search Ads

Responsive search ads are a type of listing ads that automatically combine multiple ad headlines and descriptions that have been set in advance .

Machine learning automatically tests different combinations and learns the most effective combinations. And since the optimized combination of ads is displayed for each search user, a high effect can be expected.

Advantages of using machine learning in advertising operations

The advantages of using machine learning in advertising operations are as follows.

  • Leads to an increase in the number of CVs
  • Can reduce CPA
  • You can reduce man-hours for advertising operations

Leads to an increase in the number of CVs

By using machine learning in advertising operations, it becomes easier to increase the number of CVs.

By using machine learning, based on the large amount of data accumulated in the past, it will automatically judge ” what kind of creative is best ” and ” which layer should be targeted “.

Creatives and distribution destinations can be optimized based on this result , so it is possible to increase the number of CVs more efficiently than manual publication.

Can reduce CPA

Using machine learning increases the efficiency of advertising operations and increases the possibility of reducing CPA. For example, if you use Smart Auto Bidding, the bid price will be automatically adjusted to the optimum price, so you can eliminate wasted advertising expenses .

Also, if machine learning can maximize the number of CVs , it will eventually lead to a reduction in CPA.

You can reduce man-hours for advertising operations

By using machine learning, you can significantly reduce the man-hours spent on advertising operations. For example , when using smart ads , it is possible to automate the creation of ad creatives and targeting settings, reducing the effort required for manual settings .

In addition, Smart Auto Bidding can automatically adjust the bid unit price as long as you set a budget, and reports are also available, so you can reduce the man-hours required for analysis .

Disadvantages of using machine learning for advertising operations

There are two disadvantages to using machine learning in advertising operations:

  • Data analysis takes time
  • Poor accuracy due to lack of data

Data analysis takes time

Machine learning requires a lot of data, and collecting that data takes time. In general, it takes 2-3 weeks to accumulate and analyze data, and it is said that it takes about 3 months for machine learning to be completed and stable distribution to be possible.

Depending on the content of the training data, it may take even longer in some cases.

Poor accuracy due to lack of data

Machine learning enables more accurate predictions, settings, and advertisement distribution by learning large amounts of data . Therefore, the lack of data required for training can lead to a decrease in accuracy.

Note that Google Ads requires at least 30 CV data in the past 30 days to improve accuracy .

Three points to keep in mind in advertising operations using machine learning

In advertising operations using machine learning , it is necessary to keep the following three points in mind.

  • Minimize number of ad groups
  • Limit the number of ads in one ad group to about 3
  • Add broad match keywords

Minimize number of ad groups

Minimize the number of ad groups in ad operations using machine learning . Too many ad groups will spread the data, and machine learning algorithms will take longer to collect the data .

It is recommended to create campaigns for each product/service and create ad groups for different targets in order to prevent data dispersion .

Limit the number of ads in one ad group to about 3

It is also important to prepare about 3 patterns of advertisements for each ad group .

Smart advertisements using machine learning create and deliver advertisements that are expected to be more effective for users’ search keywords and interests .

If you prepare multiple ads , the range of options will expand , so it will be easier to deliver ads that can appeal more effectively .

Add broad match keywords

To facilitate machine learning, add keywords with a match type of Broad Match.

Ads will now appear for searches related to the keywords you specify , allowing you to gather the data you need for machine learning faster .

Also, by displaying ads for a wide range of keywords, you can expect an increase in the number of CVs .

In-house or outsourced advertising?

What is the ideal relationship between advertisers and agencies? We will show you in what cases in-house or outsourcing is appropriate.

Two successful cases of advertising operations using machine learning

Here, we introduce two successful cases of advertising operations using machine learning.

69% increase in CVs with the introduction of automatic bidding and partial match

Tosho Trading Co., Ltd. develops multiple Internet media in various industries such as finance, human resources, and education . One of the media, “Career Guide,” has been operating with manual bidding and perfect match, but the number of CVs and cost effectiveness have been sluggish.

Therefore, by introducing automatic bidding and using both exact match and partial match, we succeeded in increasing the number of CVs by 69% while maintaining cost effectiveness .

Achieved a 24% increase in CVs and a 13% reduction in CPA by utilizing machine learning

Union Eternity Co., Ltd., which provides a scrap car purchase service, was working to acquire CVs by manually setting the unit price per click, but while conducting large-scale operations, they felt that unit price adjustment was an issue.

Therefore, we switched to “maximize conversion value”, a bidding strategy of smart automatic bidding, and achieved a 24% increase in CV and a 13% decrease in CPA .

This is an example of how the promotion of machine learning led to an increase in CV while reducing man-hours .

Common Q&A on machine learning in advertising

Finally, we’ve rounded up some common questions about machine learning in advertising .

What is the learning period for listing ads?

The learning period for listing ads by machine learning is usually said to take 2 to 3 weeks , but it depends on the number of ads placed, the number of clicks, and the competitive situation. According to Google , factors that affect the length of the study period include:

  • campaign
  • ad group
  • keyword
  • Number of CVs
  • Duration of the conversion cycle
  • bidding strategy

Try not to change the learning period as much as possible, as changing the settings may adversely affect your learning.

What does “learning” mean in a bid strategy?

One of the Google Ads bid strategy statuses, “Learning,” indicates that the campaign is in the process of optimizing the bid for the campaign.

Mainly, when a bid strategy is changed , it is often under learning, and Google ‘s official help lists the following four reasons for being under learning.

  • Is a newly created or reactivated bid strategy
  • Changed bid strategy settings
  • Added/deleted/changed conversion actions related to bid strategies
  • Adding or deleting campaigns , ad groups, or keywords

Keep in mind that while it’s learning, there may be slight fluctuations in performance.

How long does Google Ads ‘learning’ last?

If your Google Ads status is Learning, it may take up to 2-3 weeks for your bid strategy to adjust to your new goals. However, it may be shortened depending on the amount of conversion data.

Use machine learning to optimize ad operations

Ad operations that utilize machine learning can be expected to bring benefits such as an increase in the number of CVs, a reduction in CPA, and a reduction in operation man-hours.

Leave a Comment