The mainstream population probably first became aware of machine learning in 2011, when IBM's Watson beat Ken Jennings, the longest-reigning human champion on the game show Jeopardy.
Watson studied content from Wikipedia, had access to 200 million pages of structured and unstructured content, and was taught basic linguistics so it could figure out how to phrase a question as an answer and the answer as a question. If you think about it, it's a lot more complicated than the Turing test, where all the machine has to do is answer questions like a human. Apple's Siri, Google Assistant, Microsoft's Cortana, and Amazon's Alexa are present-day implementations of the same idea, with even more diverse sets of data to study.
Smartly and relevantly leveraging data in all its shapes, sizes and types is a challenge every agency faces. In analyzing the explosion of consumer generated content and other phenomena like photo/video sharing we have an incredible opportunity to figure out what makes consumers tick, so we can help our clients connect with them as efficiently and effectively as possible.
The impact that integrating machine learning can have on the whole process is pretty staggering. Epsilon's agency discipline has come to regard machine learning as a critical business accelerant, because it has helped us move the digital transformation of our clients' businesses forward faster than ever before possible.
To list just a few examples, machine learning can:
Radically expand the scale and scope of market research
Like social listening, the machines get feedback and commentary straight from consumers -- they are not answering questions we posed, they are expressing themselves on topics of interest. The machines' separate expertise is bringing entire categories into an analytical context that allows us to closely examine the key consumer dynamics at work -- competitors, products, attributes, occasions and perceptions, as well as the consumers themselves. After years spent in traditional insight work it's disorienting -- even a little disturbing - to contemplate so much of the world from a true bottom-up perspective. You never know exactly what you're going to get.
Be an insight accelerant
No human brain can unpack a conversation that includes 250,000 Facebook posts. Great quantitative researchers, industry experts and the most dedicated qualitative researchers in the world might never notice or think to ask about the patterns machine learning can uncover. What this really means is that subsequent to machine learning studies, they can all use custom research to dig deeper from much more accurate and interesting points of departure. Of course, machine learning is unlikely to ever fully answer the "why?" questions…
Help your campaigns work harder
Traditional A:B copy testing becomes more interesting and effective when it evolves to A:Infinity testing in real-time.
In other words, the right kinds of machine learning applied in the right ways can give agencies and brands a permanent -- well, make that semi-permanent -- head start. Once the machine figures out who's talking about what and how, and what the major dynamics are in the conversation, we can cut closer to more relevant chases faster.
If I'm trying to differentiate my minivan brand and I already know that minivans are not only for self-employed people, but also for couples partying, I might look at innovation or targeting a little differently, not to mention creative development. This is not your father's social listening.
Combining a machine learning company's capabilities with an agency or brand's proprietary data assets is a uniquely powerful way to shape communications for our clients now, with an eye toward designing ever more immersive consumer experiences in the future. Keep in mind that from the agency perspective, this is still a fairly new technology, so the takeaways and capabilities are still evolving. How are you using machine learning to power your efforts?