AdTech Today: A.I. versus the Human Component

With the rise in computing power and the emergence of automation in industries including healthcare, retail, finance, automotive and advertising there are a constant flow of incremental updates, user improvements and partial solutions to ease workloads, improve efficiency and capture consumer interest. Algorithms and machine learning are making leaps and bounds pushing automation into new territories. But what is happening behind the curtain? Stepping back for one moment to quote Aldous Huxley,

“Science and technology would be used as though, like the Sabbath, they had been made for man, not (as at present and still more so in the Brave New World) as though man were to be adapted and enslaved to them.”
-Brave New World (First Perennial Library Edition Published 1969)

Is a human beholden to data and automation or does the inverse in fact hold true? The human component should not be underestimated and is essential in not only a philosophical context but also in any strategically implemented and automated marketing campaign.

So where should human oversight occur? What is left for a human after final negotiations have been reached and the marketing plan agreed upon? There are intuitive decisions at all stages requiring human intervention based on shifting inputs that even advanced computation cannot (yet) address.

Welcome to campaign optimization.

Optimizing is arguably the most time consuming step in running a successful programmatic advertising campaign if the advertiser, agency and client have undertaken adequate due diligence. During pre-campaign planning there should be effective, well-defined and clearly communicated parameters and goals established between the agency or in-house team and the client that outline and reinforce the overall brand marketing strategy and the specific campaign.

Programmatic Campaign Planning

A typical programmatic campaign is broken down into the following four phases allowing for further examination of human and artificial components at each successive stage.

Programmatic Campaign Outline (Planning Phase) : HUMAN

  • Targeted Site List
  • Ad Context (Ad Types)
  • Audience Data (Contextual Targeting, Keyword Targeting)
  • Safety

Campaign Initiation (Startup Phase) : HUMAN/A.I.

  • Tag Testing
  • Remarketing Pixels
  • Resource Caps and Parameters (Campaign Daily Budget Spend, Flighting, Frequency)

Campaign Strategy (Implementation Phase) : HUMAN/A.I.

  • Execute Media Buy
  • Install Testing Opportunities for Optimization (Granular Targeting Approach)

Optimization (Testing Phase) : HUMAN

  • Analyze Successful Testing Strategies
  • Evaluate Conversion Metrics (eCPM, CTR, Viewability)
  • Identify Aggregate Reporting
  • Assess Granular Performance Data (Site-Specific, Creative and Viewability)

A healthy mix of human driven insight and testing oversight with regard to automated advertising strategies determines campaign success or failure. Combining artificial intelligence tools with human intervention leads to an efficient, performance driven and highly optimized advertising campaign. Automated tools are capable of processing huge volumes of data without a break. A human can ditch the redundancy and inefficiency associated with computation and instead guide the campaign through each phase adjusting for errors, integrating anecdotal knowledge and applying other insights. These small but significant adjustments are integral in ensuring the campaign maintains the correct course.

Agencies already understand and brands are learning that a silver bullet solution in the form of a full-stack, single platform end-to-end system does not yet exist. Stringing together different platforms still remains the only way to synthesize a variety of different tasks including data collection, ad buying, audience identity mapping, targeting, remarketing and ad delivery execution.


Filtering smaller sub-groups from lookalike audience samples allows artificial intelligence platforms such as Adgorithms’ Albert to iterate small, relatively cheap campaigns. This process known as micro-segmenting enables these simultaneous mini campaigns to yield data that in turn allows a system like Albert to predict different ad elements in A/B testing scenarios to select the best performing combinations. The most successful end result(s) are deployed at larger scale achieving a dramatic increase in conversion opportunities.

Lookalike audiences are generated from isolated user profiles developed from first-party (CRM) data representing the highest-value sample of past customers. This “high” value should be attributed to customers that exhibit the appropriately defined performance metrics as outlined in the planning stage (e.g. leads, keywords, ROAS, etc.) This process represents another phase requiring human participation and evaluation of collected data samples.

The true value of applied technology solutions improves efficiency and precision at high volumes (millions per minute) of consumer touchpoints occurring daily through online and offline user activity via messaging, survey outreach, social channel engagement and website browsing. The adjustments that synchronize setup, execution, testing and optimization require human inputs to adequately bridge gaps a computer is unable to sidestep in delivering a successfully run campaign.

Ian Herman is a digital advertising and marketing specialist with a background in programmatic (automated, data driven, real-time marketing). Read more on Ian’s programmatic advertising blog AxialX