Artificial intelligence (AI) and machine learning algorithms are mainstreaming in a way that was never before possible, and these changes are having a significant influence on the way in which marketers need to approach search advertising.
In addition to AdWords itself incorporating AI into its framework, new opportunities are arising that can give marketers an edge over their competitors, or automate lower-level tasks, freeing up more time for strategy.
Here are four ways you can start taking advantage of AI to make the most of your AdWords campaigns.
Automated bidding
Automated machine learning as a solution to the decision of what price to bid on paid advertising is becoming an increasingly popular option as the necessary technologies become available to more firms.
Bidding too low means missing out on opportunities to reach leads, while bidding too high means sacrificing ROI.
Google’s internal automated bidding, on top of being identical to what everybody else is using, doesn’t have access to the information it needs in order to maximize your ROI. Reaching that goal also requires knowing consumer trends, purchase behavior, seasonality, demographics, customer lifetime value, and more.
A successful automated bidding model must:
- Estimate the price elasticity of each ad by using statistical inference based on previous bids
- Factor in the actual value expected by a click from each individual ad based on previous clicks
- Iterate in response to new data
- Recognize changes in the bidding landscape or the performance of visits and adapt quickly, rather than falsely assuming past performance will predict future performance in all circumstances.
There are, however, some things to look out for:
- Models that don’t know what’s happening on your site will make bad inferences. For example, if you test a new landing page and it turns out to lower your ROI, your model could start bidding lower on those keywords. After replacing the landing page with a better one, the model may still get stuck bidding low on the keywords, because there isn’t enough new data available to push the bids back up
- Models that rely too heavily on statistical significance may test a losing strategy for too long, but models that fail to incorporate statistical significance can throw away good opportunities while propping up flukes.
- Watch out for feedback loops in your model. For example, you wouldn’t want your model to bid more on an ad with a high conversion rate if the only reason the conversion rate is high is because the high bids are increasing the conversion rate. These types of conflicts should be controlled for.
Pausing poorly performing ads
The quickest way to lose money in AdWords is to continue bidding on an ad that isn’t producing any ROI. When the clicks roll in but the sales don’t, this can be a disaster.
Similarly, when an ad is getting the bids but not the clicks, your quality score will suffer, and ultimately your ROI will follow suit.
A well-built machine learning algorithm will understand when it is necessary to pause an ad in order to avoid hurting your ROI or quality score.
Here are some important considerations your model must account for:
- The model must not be so sensitive that it abandons ads before they have a chance to show ROI. It must use statistical inference to estimate potential losses and gains based on previous performance
- Rather than pausing the full ad outright, the model should factor in individual segments that can be paused, such as traffic from mobile devices, certain browsers that are not producing revenue, times of day or days of the week that repeatedly do poorly, or ad variations that aren’t performing well.
Dynamic ads
AdWords’ Dynamic Search Ads are one piece of machine learning technology that currently come built-in with the platform, allowing anybody who is using AdWords to take advantage of it.
Dynamic Search Ads automatically generate headlines to capture a searcher’s attention. After uploading a list of landing pages that you want Google to generate dynamic ads for, Google will identify searches that are a good fit for your landing pages, then automatically generate ad content using phrases from your pages.
Google is also generating ad suggestions based on machine learning. These recommendations use models of prior performance to suggest changes to your ads that should boost your results.
But the possibilities for dynamic ads don’t end with what is native to AdWords.
Machine learning approaches can be used to create dynamic ad content that incorporates the following:
- Mixing and matching copy, image, and audience with multivariate testing and evolutionary algorithms
- Incorporating the influence of external factors such as the weather or time of day.
A few platforms experimenting with this kind of control include Sentient Ascend, IBM Watson, Zalster, and Refuel4.
Available platforms
The previous insights might make it sound like you’ll need data scientists and developers on your team in order to take advantage of what AI and machine learning have to offer, but this isn’t necessarily the case. While full-time dedicated AI staff are a good idea for big businesses, small and medium businesses can still take advantage of these emerging technologies with emerging products.
Here are just a few examples:
- Acquisio: This machine-learning platform is designed to improve performance in AdWords, Bing, and Facebook ads by cutting CPC and CPA while raising clicks and conversions
- Cognitiv: Uses deep learning to predict where best to spend your money, self-customizing for each brand based on historical data
- Frank: In addition to AdWords and Facebook ads, Frank is connected to millions of publishers. It launches campaigns automatically and optimizes them by target audience, creative, and channel
- Magnetic: Designed to automatically match audiences to inventory while optimizing bids and cracking down on fraudulent clicks
- Quarizmi: One of few AI platforms that specifically bills itself as being for AdWords. The platform automates keyword discovery, creative, bids and campaigns
- Trapica: Identifies audiences, matches them to creatives, optimizes bidding, and scales your campaigns.
No matter the platform, use the insights discussed to make informed decisions about what will work best for you.
Conclusions
As AI becomes mainstream within the PPC industry, marketers will need to begin shifting their areas of expertise away from micromanaging keywords and bid prices, and towards higher-level strategy. In the meantime, the techniques and platforms discussed still aren’t in use by the majority of your competitors, and taking advantage of that gap would be a wise move.