Programmatic Optimization: Do Machines Trump Humans?


Advertisers often don’t know what’s happening behind the scenes of a programmatic buy. Many think programmatic buying uses machines with fancy algorithms to target specific audiences, maximize reach, convert audiences and more. This is true, but what are the human elements? What happens after the buyer hits the “start” button? Is that when the technology takes over?

 Myth: Once a programmatic buyer hits the “start campaign” button, she lets the campaign run and allows the machines to optimize.

Fact: Successful campaigns, even programmatic ones, require human intervention.

Human Optimization vs. Machine Optimization
When reviewing audience targeting during the campaign, algorithms will recommend behavioral audiences to target based on look-alike modeling (similar audiences to those already converting). However, it’s important that humans evaluate those audiences based on size, cost, overall performance and brand objectives. Below is an example of how people do this during a campaign.

Current scenario: The agency purchased a group called “Soccer Moms” from Vendor A at a 50-cent cost per thousand (CPM) with an audience size of 26 million and a low cost per action (CPA).

Machine-recommended optimization: The machine recommended purchasing “Soccer Moms” from Vendor B at a $1.00 CPM with an audience size of 7 million.

While it’s great that the machine recommended purchasing “Soccer Moms” from Vendor B, a human should evaluate the risk. In this example, he should evaluate a few things:

  • Determine if Vendor A has reached its total audience. If not, it does not make sense to open targeting.
  • If the campaign has reached the total Vendor A audience, the buyer may consider Vendor B to add reach.
  • Finally, a buyer should evaluate the risk of adding Vendor B audience and determine if it will increase CPA.

The Value of Human Judgment
The machine automatically bids for the right impression, at the right time, to target the right consumer. Conversely, it’s imperative for the human to evaluate what categories and sites the machine is targeting based on the brand’s objectives and guidelines. She should opt out of categories that simply don’t make sense for the brand. For example, a financial client may not want to show up on categories or sites with political or gossip content. This also includes monitoring for brand safety and fraud. Or, a human can create a best-performing site and category list to increase brand presence on sites and categories where performance is high.

Finally, human intervention is necessary when machines automatically optimize into the campaign goal. For example, in a campaign where the goal is CPA, and retargeting has the lowest CPA, the machine might suggest optimizing all dollars into retargeting. However, it’s key to remember that prospecting drives people into the retargeting pool. A human should ensure that not all budgets and impressions are pulled into retargeting, and the campaign maintains efficient levels of prospecting.

It’s essential that humans stay involved in programmatic buys and don’t allow machines to do all the work. Campaigns perform much better when a person is optimizing and monitoring in real time. Additionally, sharing optimizations, audiences, targeting and learnings with brands is important to the performance of future campaigns.

Want more content like this? Subscribe now!

Lindsay Pullins
Lindsay Pullins

Lindsay lives and breathes digital marketing technology, specifically programmatic buying. Her passion for programmatic comes from her desire to uncover new insights and understand the buying behaviors of online consumers. These learnings allow her to customize successful programmatic campaigns for Empower clients. Lindsay’s knowledge and experience helps drive superior results in the world of automated ad buying.