The recent Knowledge Discovery and Data Mining conference in NYC this past week gave me a good look inside the belly of the data beast and I was humbled. I realized that the world of data science was moving faster than we in the media industry realize and the challenge that faces us now is insuring that our data systems and protocols keep up with the these new advancements.
In fact, data science is being applied in a range of pro-social efforts today and the KDD conference highlighted many worthy efforts in support of people and society. Microsoft’s Eric Horvitz, for example, spoke of how data is being used in transportation – monitoring wind patterns to help lower the carbon footprint and mitigate the impact of storms as well as monitoring cell phone usage in under developed countries to help pinpoint natural disasters like earthquakes to facilitate aid. Data science is also expanding in medicine to lower hospital infection rates and reduce inpatient recidivism. Horvitz explained, "The value of data is to increase and enhance your decision making" no matter what industry you target.
One aspect of new data analysis that struck me as applicable to the media industry is “rich representation” which enables the user to dynamically tag elements in a video. It involves some facets of facial recognition, body parts, clothing, items, landscape identification and other features. In this way a video can be more easily categorized by elements in the content. This capability will enable content owners to more fully and accurately categorize the elements in their content and might even enable a more granular way to measure small but discernible facets of content for performance success.
Another analytical application for media is real time speech translation which has improved dramatically in the past 5 years. It is now possible to translate conversational speech in real time, even Skype to Skype, which opens up possibilities for faster global distribution of content.
Further, applications like adaptive diversity (a form of data mashing), transductive learning (a type of machine learning), consensus modeling (used for mining data to optimize group recommendations) and collaborative filtering (used in recommendation engines) can be applied to media content selection in a variety of ways; performance prediction models, program scheduling that enhances audience flow, recommended content selection by viewing segments and the ability to create and refine those segments.
Some prescient companies like Bloomberg.com are using data science to construct custom consumer segmentations using a disparate selection of data sets including, according to VP Technology Pat Moore, the origination network, device use, traffic flows which are then used to create graph models to match users with similar features for a specific advertiser.
View the short interview with Moore here:
Claudia Perlich, Chief Scientist at Dstillery, uses data science in consumer targeting, seemingly getting to one-to-one marketing. She explains, “I develop algorithms that utilize data to make marketing more focused and ultimately more effective for our clients. Specifically, I apply machine learning and predictive modeling techniques to distill billions of individual events of consumer behavior into an audience of prospective customers. Every day, we analyze billions of data points generated from where people go on their devices and with their devices. Instead of trying to bucket people into demographic or behavioral groups, we evaluate every consumer individually with respect to this specific sequence of actions to detect potential product interest and then identify the precise moments and channels for a brand’s message. We buy an impression only when we know the consumer is likely to engage. This allows us to be incredibly selective …. While others bid on 45% of impression inventory, we bid on only 3%. This approach is individualized to every brand or product, and it’s also individualized to every consumer on the other end.”
I suspect that this is just the beginning. And unless we as an industry consistently make an effort to understand the expanding capabilities of data science - machine learning, artificial intelligence and data mashing for example- we will fall short of optimizing our data for viewership, cross platform, POS and ROI measurement uses. Of course data intelligence, like everything else in our industry today, is evolving quickly but we should at least begin to include the basics of data science in our media and marketing research conversations so we are not leading from behind.