Marketers are a curious people, said Yoram Greener, Partner, Technology, Digital and Social Analytics, JubaPlus, a marketing analytics consultancy, in a recent DMA Analytics Council webinar. “We want to know as much as we can about our customers and prospects. Marketers have been struggling to keep up with the increased and multi-channel consumer interactions, to master the technology options and understand consumer behavior with insights that apply to specific business needs.
“The goal is to turn away from what we THINK customers want and need. We must truly understand the truth of their motivations – and machine learning is an analytics key to gaining that understanding,” he said. Yoram presented “Will Machine Learning Algorithms Transform Real Time Digital and Social Interactions?” as part of the DMA Analytics Council presents webinar series.
Getting to that truth is not an academic exercise, it’s very practical, he said. “It all starts with the customer, the greatest asset for any brand. No matter how you map out your customer journey – a traditional linear funnel or a circular path – every person has his own journey and accesses information across different devices.
“It’s a very complex environment with a large influx of information,” he acknowledged. He suggested that success will be found in capturing and understanding specific “moments of interest.”
“Those marketers who are quick to react because they were quick to capture the insights from raw data will do better than others, “he said.
Consumers, of course, don’t think about “moments of interest.” They search and make decisions – anywhere and everywhere, Yoram said. Marketers who track these “moments of interest” create intersection points where audiences can be guided toward profitable interactions. “If 60% of my visitors are coming and leaving in 10 seconds, then we need to maximize the information provided in the 10 seconds,” Yoram said.
“We call these people ‘passers’ because they come and go and we do not know their motivation. It gets interesting if some of them will come back and spend more time,” he said. “We try to understand the intention of those consumers –and this can generate a very wide set of answers.”
By mapping these, and knowing you will be wrong on some of them, we will convert some at “leaps and bounds,” he said. The testing developed to optimize the conversion is the work after the algorithm reveals patterns and profiles.
“It’s not just the machine but machine learning – continuous learning. To learn down at the visitor level, what are those intersection points, and how people are reacting,” he said.
Some of the big marketing challenges that can be addressed through machine learning algorithms include:
Industry has reacted effectively – there are many tools out there to help with web personalization, social media management and search bidding and yield management, Yoram said. While these help marketers cope with the challenges, many are not yet holistic, efficient solutions, he advised.
One technical advancement to address this challenge is the emergence of DMPs – data management platforms – that can take masses of very diversified data and let the analysts interrogate the data on several levels. For example, there are a few companies like Akamai which successfully cope with this challenge of cross device cookie assigned to a single user, he said. The DMP is not “a theoretical wish list, it is something that every digital marketer of today needs to allocate budget to,” Yoram said. Those that are worthwhile can be at least 20% customized to be specific to the brand architecture and need, he advised.
“The only way we can satisfy consumers is to improve the speed, timing of reaction and relevancy of content in our marketing. Algorithmic solutions are the answers. They can address these dimensions at the same time,” he said.
Most marketers have goals around bringing more of the high value customers to our website and spending the right amount of money to get them there. Equally, we need to know who is of low value. “There are margins of error, algorithms are not the new God, but despite those errors of judgments, we will be much better off,” Yoram said.
The process of developing algorithms is rigorous, he said. “We deal with a lot of data (both rows of records and columns attributes) and a lot of matching. A typical environment can be millions of rows and thousands of columns.
“We take data from different systems, so it has to be normalized. Once it’s normalized, we can start the mining process – which is the most interesting part. It’s a bit like searching the ocean for some signal you can anchor on. This hard work is what you should expect from your analytics team,” he said.
Once you understand the customer at an individual level, then you can improve the path to conversion. And, isn’t that the goal every marketing organization?
## ## ##
Editor’s note: This webinar is one of six presented this year by the DMA Analytics Council. Check out the full schedule and register now. More information on the Analytics Council and our activities is available here. Many DMA memberships include Council participation – check with your account manager or email Stephanie Miller.