Google AdWords Machine Learning Updates: What You Need to Know
One of the key initiatives that Google has outlined for their AdWords platform in 2017 is the implementation of machine learning. The ability for their system to better analyze and recommend solutions to their advertisers presents a huge opportunity.
The concept of machine learning has been around for many years, but technological hurdles have kept it from mainstream applications. However, in recent years we have seen a sharp incline in momentum for machine learning. Simply put, machine learning is a method of data analysis that automates analytical model building. It is a branch of artificial intelligence, based on the idea that machines should be able to analyze data to learn and adapt through experience. The potential for machine learning solutions in business applications is endless.
Why is that important?
In the increasingly complex digital ecosystem, we are able to create our own models that analyze large amounts of complex data. This enables businesses to deliver faster, more accurate results on a large scale. It is especially relevant for marketers, who try to reach potential customers at the exact right moment in their buyer journey.
The designing of these models takes a lot of time and manpower. Having machine learning there to assist with building these models gives companies of all sizes a huge advantage. With the amount of data generated by AdWords advertising campaigns, you’d be hard-pressed to find a better testing ground for machine learning technology.
Google has announced several machine learning additions to their AdWords platform. Let’s dig into these machine learning features and explore how it will affect the AdWords platform.
One of the most visible machine learning features that Google implemented in 2017 was the Conversion Attribution feature. The feature deals with how conversions are attributed to the various channels and touchpoints throughout your marketing funnel. Using several different modeling methods, Google allows marketers to decide how the credit for conversions is applied to multi-channel marketing efforts.
The inclusion of Attribution in the Adwords system makes it easier for marketers to measure impact across multiple channels. It attempts to provide systems that allow you to better measure the impact of mid-funnel marketing materials. Let’s take a look at some of the different Attribution types that Google currently offers:
- Last click attribution. Conversions are attributed to the last channel that the user clicked before converting.
- First click attribution. Conversions are attributed to the first channel that the user clicked before conversion.
- Linear attribution. Each channel that the user interacts with prior to conversion receives equal credit.
- Time decay attribution. The most recent touch in the conversion path receives the most credit. Previous steps receive some reduced credit for the conversion. The more time between touches, the less credit each individual touch receives.
- Position-based attribution. First and last touches each receive 40% of the conversion credit. The remaining touches split the final 20% of the credit.
- Data-driven attribution. Google’s own algorithm uses data from your account to assign the credit for the conversion. The system is capable of assigning credit to specific ads, keywords, or campaigns.
Google Attribution runs these models and feeds that data back into AdWords. There, companies can use the data to implement a flexible bidding strategy or use the data for optimization. Attribution will help companies better allocate their advertising funds, and improve ad optimization.
Custom In-Market Audiences For Search
Another machine learning feature that Google has announced is Custom In-Market Audiences for Search. The system will allow companies to reach customers that are actively researching and comparing products that they offer. Google identifies users with buying intent and gives advertisers the ability to target them directly.
Essentially, custom-in market audiences are an additional layer of targeting on top of the traditional targeting methods in AdWords. They can be used in conjunction with targeting options like keywords, location, time of day, etc.
In-market audiences are different from other interest categories because Google has identified their intent to purchase. There are a number of ways that Google analyzes user activity to determine that they are getting ready to buy, including:
- Content of websites visited. What kind of content are they accessing? What does that content say about their intent? The length of time that users stay on these sites and interact with the content may also play a role.
- The frequency of visits. Has the user seen a recent uptick in visits to websites that would indicate intent? This is a sign that they are ramping up their product research before making a purchase.
- Account clicks on other related ads. Has the user been interacting with other ads in the same category? This is common during the researching phase as users reach a buying decision.
- Subsequent conversions. Has the user converted for other advertisers in your market? A conversion for a complimentary product is a strong signal for buying intent.
In short, Google uses a number of metrics and data points to determine that a user is actively looking to make a purchase within your market. Custom in-market audiences debuted in 2013 under the name “In-market buyers,” for a short testing period, and is now available to all users in more than a dozen market categories.
Life Event Targeting
Life Event Targeting is another interesting machine learning feature in the AdWords platform. The feature appeared on the Facebook Ads platform, and Google quickly followed suit on AdWords. Essentially, Life Event Targeting looks at patterns of behavior that indicate an impending life event is on the horizon. Search queries that suggest that the user was looking for a new apartment, researching venues for an upcoming wedding, or looking for a new job would all be good examples of life events that could be targeted.
Google currently allows life event targeting on YouTube and Gmail. The feature is not currently available in other ad formats, but it is a safe bet that Google will someday look to expand life events to search and other Google-branded platforms. Life event targeting presents a lot of potential in creative advertising. During the Google Marketing Next event, Google highlighted Sonos as an example of a company that saw amazing results from the life events system. Using the feature, they were able to generate some pretty impressive results:
- 50 percent lift in ad recall
- 35 percent lift in purchase intent
- 5x lift in people searching for Sonos on Google
While not every company should expect these kind of results, it goes to show just how much power there is in this kind of targeting.
Smart Display Campaigns
Smart display campaigns are another exciting machine learning addition to the AdWords platform. In fact, the feature may open the door to the most hands-off method of advertising the industry has seen. Smart display campaigns allow you to quickly deploy new ads and campaigns with a few clicks of the mouse. They are generally great for targeting customers that are early in the buying decision process. Using Smart Display, AdWords does all of the heavy lifting by handling bidding, targeting, and even the creation of ad creatives.
In the process of creating a traditional display ad, you’d be prompted to upload a completed ad image for the display network. Using Smart Display, Adwords will ask for a few different assets and use those assets to automatically create its own responsive ads for testing within your campaign. Additionally, Smart Display can scan your website for images that it can incorporate into your ads.
This is a great feature because it makes it easy to auto-generate branded ad images without waiting for designers. While the images themselves don’t always turn out perfect, it is a great solution for companies that have trouble putting together enough creative images to test within their campaigns.
Once the ads have been created, you do have the chance to approve or reject individual ads to ensure that no auto-generated ads run that may be damaging to your brand. Google reports back with the performance of each asset, assigning scores of “Best,” “Good,” or “Low.”
For business owners with small marketing teams that prefer to stay lean, Smart Display could be an excellent tool in their arsenal. The system offers a simple, intelligent solution to managing the complex variables of display advertising. Additionally, the system is able to show ads in nearly any format across the entirety of the display network. If you are struggling to reach your max advertising budget, Smart Display could be the most effortless way to broaden your customer base and win new conversions.
Machine learning is certain to play a huge role in the future of digital advertising and the Google AdWords platform. It is clear that Google has placed increased priority on bringing machine learning features to the AdWords platform in 2017, and we can expect to see this trend continue as Google continues to flesh out the features they’ve already introduced, while continuing to bring new features to the platform. Machine learning represents huge potential in the digital advertising industry, where small optimizations can mean tens of thousands in new revenue for businesses.