Promise and pitfalls of machine learning in advertising outlined at recent PHD event

Machine learning is revolutionising advertising. However, businesses need to make sure that machine based optimisation drives growth, and not efficiency at the expense of diminishing revenue.

This was the view of Konrad Feldman (left), global CEO of Quantcast, on the future of machine learning and its effect on advertising at an exclusive lunch for PHD clients in Auckland on Monday.

Feldman said machines are goal oriented and want to be rewarded.  So left unchecked, they will put performance ahead of profit.

Says Feldman: “If we measured advertising campaigns on actual incremental outcomes, we wouldn’t worry about fraud because bots don’t buy anything.  But we’re not measuring on incremental outcomes, we’re working on a proxy – clicks or last touch attribution – and that proxy is easy to gain.”

PHD’s group general manager digital, Christophe Spencer, says big data is a valuable tool, but it needs expert navigation. 

Says Spencer: “Machine Learning presents a huge opportunity, giving us the resource to process, analyse and interpret the vast amounts of rich advertising and consumer data at our disposal. This in turn helps us deliver more targeted, informed and relevant messages to consumers. However our fundamental belief is that people must play a central role alongside machines, providing context and delivering a deeper understanding and the insight in the data. Our decision to bolster our team with the addition of specialist data analysts reflects this approach.”

Feldman co-founded Quantcast in 2006 to transform the effectiveness of online advertising through science and scalable computing.

Driven by increased computing capacity and the huge amount of data companies can now collect, he expects Machine Learning to continue to strongly influence advertising decisions.

Says Feldman: “The ability to understand and interpret data is something that’s always been around, but we’re doing it at new scale. If we’re going to have increasingly used computer systems that are goal driven, how can we set them up for success?”