Will Machine Learning Make Performance Marketers Obsolete?

Before we start, some context will help you understand why I have this question. I have spent significant time in performance marketing during my career. I still remember when I started, Google Adwords was completely different from what it is today. For example, there was a rule that if your keyword click-through rate was below 0.5%, your keyword would become “inactive” and it would be extremely difficult (or impossible) to recover an inactive keyword.

Facebook was founded in the same year, so there was no Facebook advertising yet. 😀

Fast forward to 2022, machine learning is at the center of both Google and Meta advertising platforms (and many others like Amazon’s). Both companies are advocating for account setup simplification (which gives the machine more data to work with), a diverse set of creatives (both formats and concepts), and of course, site tagging/conversion API to give the machine the outcome signal (aka conversion) that advertisers care about.

Gone is the day we need to set up a granular search campaign structure or display campaign structure to customize each audience segment with relevant creatives. Now, instead, we are advised to set up just one campaign (in the case of Performance Max for Google), and the machine will automatically find the right inventory source (search, youtube, Gmail, etc.) and serve the best creatives to the audience (via responsive ad format). Bid optimization will happen automatically via suitable off-the-shelf bid strategies each platform offers. Budget optimization between campaigns can happen semi-automatically too.

So what then do we do all day? 😀 running excel reports? 😛

Machine learning is just a tool

Yes, it is a potent tool but, ultimately, a tool. That means the machine doesn’t know what is good for your business. (A quick pause here).

The machine is incredible at achieving the outcome (conversion or ROI) that you set at the right efficiency level. However, it doesn’t know if achieving that goal suits your business.

It doesn’t know if you should run Google Ads or Meta ads, or any advertising in the first place.

The machine doesn’t know a lot.

• It doesn’t know your brands or your potential customers.
• It doesn’t know why your potential customers choose your brand over the competition.
  • It can’t create engaging messaging or landing page experiences for your potential customers without a great deal of training data.
• It doesn’t know if running advertising brings incremental revenue to your business.
  • Incremental revenue is revenue that would not materialize if you don’t run the ads.
  • Potentially, the machine may know the incremental revenue or conversion from a single channel but not at an overall level for your business.

Walled gardens limit the machine.

It can do an incredible job of optimizing within the Google ecosystem, Meta ecosystem, Amazon’s or Tiktok’s, etc… but not across them. This fact is unlikely to change soon, given the focus on consumer privacy.

This means the human decides where to run the ads and how much to spend within each walled garden.

In the next 3-5 years

So at least for now, I am not worried that my job will be replaced by a machine, not in the next 3-5 years. But I should be concerned about how to continue providing more value to the business. For example, I should continue to worry/learn more about

• Knowing the brand and its potential customers.
• Understand the power the company has or may not have.
• Does it even make sense to run advertising?
• If advertising is needed, how can we evaluate the incremental business impact of advertising on the business? The keyword here is incremental.
  • Impact on business can happen within a short period (within three months) or over an extended period (years), so we need different measurement solutions.

• How to set the machine up for success with performance marketing?
  • Given the different privacy laws and regulations worldwide, what is the framework to ensure that we respect user privacy, follow the law, and give the machine the signals it needs to succeed?
    • Gone are the days when performance marketers just collected as much data as possible and then sent them back to the ad platform.
    • Now, we need to be purposeful about why certain data is important or allowed to be collected/used. And how to send them back to the machine in the right way.
  • Creative messaging and overall user experience: diversity in formats, concepts, imagery, representation, etc… seem to be the message many platforms advocate.
• How to set the machine up for success with other non-advertising activities

What do you think?

Chandler