Businesses should consider advanced segmentation techniques by employing artificial intelligence technology.
advertisement
advertisement
The Lotto Effect
First, consider a complex “wrinkle” in customer segmentation that often goes unnoticed. Take a typical customer segmentation effort today -- a company wants to identify customer segments with significant growth potential for a particular product. To accomplish this, it captures 50 customer attributes such as age, gender, location, account tenure and other customer-level information. From this relatively small set of attributes, an analyst will need to identify the “best” six attributes to base the segmentation upon -- a surprisingly complex endeavor.
Selecting the best six out of fifty is akin to playing a lottery -- 50 numbers… pick six… 15 million ways to lose. This wrinkle quickly grows into a chasm of complexity with the impact of Big Data. In fact, a more reasonably sized dataset may have 500 attributes, which would result in over 21 trillion segmentations to evaluate. Not great lottery odds.
A new way to play the game
Traditional segmentation approaches rely on the analyst to determine the best set of attributes facing these odds, and also attempts to decipher what the “best” attributes mean. It's no
wonder that segmentations based on these approaches often lose their relevance within six months. The lottery effect, although it initially seems like an added obstacle, in fact provides the big data
landscape to search for the optimal segmentation. However, a more cutting-edge approach is required to take advantage of it, which is best addressed by artificial intelligence.
This type
of approach finds the optimal segmentation among an endless number of attributes -- and more importantly, bases it on business objectives that are embedded directly into the technology. The latter
ensures that the optimal segmentation remains aligned to the core business objectives that drove the need for it in the first place. One of the key values of leveraging artificial intelligence
technologies is that this provides a unique approach that connects advanced analytics to business goals. This is critical because objectives will vary for each company depending on their brand,
market, and priorities, which is often reflected in three core areas.
• Who to focus on -- resulting segments must be highly differentiated across core value dimensions such as
revenue, cost, profitability, NPS, and other aspects to guide investment strategies
• What to offer and what to say – each segment needs a rich, unique and comprehensive profile
to inform product, offering and messaging strategies
• Where to find them -- for the solution to be actionable, the segmentation often must be developed so that it can be applied
directly to a customer database or prospect universe for targeting purposes.
With the advent of Big Data and more advanced analytics, customer segmentation techniques based on traditional
approaches will fast become obsolete in today's omnichannel world. Not only do companies risk wasting their investment on the segmentation effort, but more importantly, they risk being misled on where
value exists in the market and among their own customers.
Jeriad, well-said. This is exactly right. Marketers face twin problems, both of which are related to the problem of audience fragmentation. First, the ability of TV to deliver mass reach -- while still strong -- is weakening significantly. We may soon need to create an alternative source of mass reach, which likely means re-aggregating audiences programmatically. Secondly, the number of segments -- and the number of places those sub-audiences might be found -- continues to multiply. The problem is literally inhuman, which is why some combination of Big Data and AI is indispensable. This is exactly what Rocket Fuel has been doing. (Disclosure: they are client.) The virtue of Artificial Intelligence in the solution is that it's advertising that learns. A fresh look at segmentation is definitely in order: the audience and the tools have evolved in ways few of us might have predicted.