The most useful definition of ‘Big Data’ that has been offered by computer science is that data sets classify as big when human analysts need to hand over to machine learning software in order to properly discern patterns in the information. Machine learning is a form of artificial intelligence in which software is written to apply an analytical methodology in an iterative manner such that it learns from its mistakes and gets better and better.
Any good pilot knows that computers are much better at doing almost all of the heavy lifting when a large passenger jet flies across continents. But no pilot steps into the cockpit of an A380 without years of training and hundreds of hours of experience. A similar logic applies in PPC. TMI introduced machine learning, or programmatic technology, to its campaign management as soon as the technology became mainstream in digital marketing. However, this introduction was predicated on the years of accumulated institutional experience that TMI had amassed via its staff manually applying and refining the techniques that would now be supervised by highly sophisticated, 24 hour-a-day software.
This repository of human intellectual capital remains a non-negotiable precursor to the deployment of any technology we use. Our PPC pilots are acutely aware that the real world of ad auctions can make a very learned machine look stupid and TMI trains its campaign managers to assume control at a moment’s notice when needed. This hybrid approach ensures that exogenous shocks do not unduly skew the information referenced by the automation tools we use to mine the huge datasets at the heart of this industry.