machine learning vs. human optimization

finding the right balance for your paid media campaigns

angela krieger
8-minute read

in brief

  • A message from Google to ad account holders explained the company is bringing in “Google Ads experts” to manage campaigns “behind the scenes.”
  • While you can (and likely should) opt out of the program, use this opportunity to revisit the capabilities and limitations of machine learning.
  • The ideal approach to campaign optimization balances the speed of machine learning with human judgment.

In 2018, marketers spent more than $600 billion globally on paid digital media across search, social media and display channels. Today’s customer expects a tailored experience and paid media enables advertisers to reach individuals at specific points in their buying cycles, so this sum is far from surprising.

In fact, Google derived 85% of its 2018 revenue from their ad network alone: $116 billion of the $136 billion total. With so much of Google’s revenue tied to its ability to reach and convert prospective customers effectively, the company has made significant efforts to boost the value they’re delivering to advertisers through machine learning.

The company has a stated goal of “putting machine learning in the hands of every advertiser.” And at a Google Marketing Live event in June 2018 a spokesman noted machine learning is “key to delivering helpful and frictionless experiences consumers expect from brands.” As true as that might be, machine learning also makes it easier for advertisers to spend on the Google network.

Google Ads account holders recently received an email message with the headline “We’ll focus on your campaigns, so you can focus on your business.” The message explained Google is bringing in their “Google Ads experts” to manage campaigns “behind the scenes.” The email raised more questions than answers.

Who are these experts?

What input do we, as advertisers, have over the optimizations these experts make?

What recourse do we have if those optimizations made by experts miss the mark?

One thing was made clear: your ad account was automatically enrolled in the program unless you manually opted out. Reading between the lines, it’s likely that Google’s machine-powered recommendation engine will bring forward opportunities, and humans at Google will implement and communicate those opportunities. The recommendation engine supports automation in ad group structure, copy creation, keyword selection, and bidding.

Asked for clarification, a spokesperson shared that Google “doesn’t guarantee or promise any particular results from implementing these changes.” That doesn’t exactly instill confidence. Instead, this announcement and approach bring to the forefront an increasingly common question I hear from marketing leaders: what’s the proper balance between machine learning and human optimization?

the benefits of machine learning

There is certainly value in employing machines for marketing efforts, and the headline of Google’s announcement touches on one of the main value propositions. Leveraging machines to do the heavy lifting can free up you and your team to focus on driving business results. While this proposition is enticing, there are flaws beneath the surface.

Before diving into the potential pitfalls, let’s focus on the benefits delivered by machine learning in paid media campaign optimization. Machines can help marketers:

  • find new addressable audiences
  • tailor keyword bids
  • deliver the right message
  • properly attribute marketing’s impact

With capabilities like dynamic search ads and dynamic ad extensions made possible by machine learning, ad platforms can serve tailored content to a specific user within search results. Used the right way, this increases the likelihood of conversion as ad formats and copy rotate automatically, increasing the odds an individual sees relevant and engaging messaging.

On the backend, smart bidding enables the platform to adjust bids in real time, focusing budget on individuals deemed more likely to convert. Advanced attribution models made possible by machine learning allow marketers to move beyond last-click conversion and better understand the impact each campaign has on an individual’s path to conversion. Together, tailored content and bid optimizations can achieve a new level of efficiency.

the power of human optimization

So, we should let the machines loose within our ad campaigns and reap the rewards while paid media strategists enjoy 4-hour work weeks, right? Not quite. In my experience, an optimization strategy based entirely on machine learning lacks the human touch required to account for important nuances. Your business is unique, your customer is unique, and your campaigns should be both unique and tied to a broader strategy.

To get the best results from machine learning, humans must play a role in assessing trends, drawing conclusions, and determining next steps. Machines can’t replace the human abilities to:

  • interpret audience insights
  • incorporate negative queries
  • tailor ad copy strategy
  • optimize for lead quality over quantity

Imagine you lead the recruitment marketing efforts at a public university we’ll call State College. Your goal at State College is to drive prospective students to apply, so you’ve asked your team to investigate which search queries drive traffic to your online application. Two queries stand out as the primary traffic drivers: “state college application” and “state college application portal.”

There’s a key difference between the two search queries, though—a difference worthy of strategic consideration. Individuals searching for your application portal are likely trying to finish their application or check its status, not begin an application. This audience should have a different experience from those who are looking to begin an application. The machine alone is unable to recognize and tailor experiences this way.

strike an appropriate balance

Guided by people who understand what makes your business and customer unique, machines play an integral role in the rapid optimization of paid media campaigns.

Left to their own devices, machines can too easily identify opportunities at odds with business goals, they can struggle to highlight key value propositions in ad copy, and they can optimize toward keywords that drive form fills and phone calls, but not revenue.

The correct balance requires a team of experts to set the strategy, ensure it’s aligned to what matters most for your business, and check in on a frequent basis to both course-correct and mine key insights. In other words, the touch required from the machine is constant, while humans can check in daily.

If you’re interested in learning more about best practices for incorporating machine learning into your paid media approach, or you’re wondering if you opted out of Google’s machine-powered program in time, feel free to reach out.


resources

Machine Learning in the Hands of Every Marketer
Total Media Spend Rising
Google’s Annual Revenue
Google Tells Advertisers…