As technology continues to evolve and we all grow more comfortable with using new digital channels to interact with family, friends, co-workers and even brands, marketing leaders find themselves faced with two complex challenges:
- how to effectively maintain necessary customer interactions, while honoring customer privacy and channel preferences, and
- how to determine when customer interactions are required – before a customer loses patience, or worse, defects to a competitor.
Looking back only four or five years ago, customers had precious few options if they wanted to communicate their dissatisfaction with a brand. They could phone the call center, send a letter, walk into a store or branch . . . or take no action other than to complain to a few friends. In today’s world, though, that dynamic has been turned upside down. Now, within minutes, customers can use social media to create whole movements against or on behalf of a brand, a company or even a government.
Given this paradigm shift, marketers need to detect brand-negative activities initiated in channels they do not own, and they need to detect them early – so they can take steps to avoid larger consequences. What’s more, detecting potential problems is just one part of the puzzle. Once identified, marketers must also predict whether a particular online “event” requires action, and more importantly, what kind of action is optimal.
Back when I started in campaign management software development 13 years ago, marketers had to capture and respond to relatively few brand-negative events, and most could be handled by creating a set of rules or manual actions, based largely on marketing experience and gut feel. However, as both the number of channels and the number of digitally-connected consumers started to skyrocket, the volume and variety of events needing detection increased, too.
At first, marketing organizations implemented rudimentary predictive analytics to automate handling brand-negative activities. But almost immediately, trouble emerged: Highly trained individuals, using tools like SAS and SPSS, developed predictive models based on each type of event activity, creating a logjam for the marketing user. Now, as the number of customer interactions and events continues to grow, those logjams are starting to swell. Short of hiring an army of predictive analytics Ph.D.’s, how can marketing organizations possibly keep up?
Related Resources from B2C
» Free Webcast: Know Your Story, Understand Your Customer
Going forward, predictive analytics can help in three main ways. The next generation of predictive analytics needs to be:
- Smarter. New predictive analytics solutions must provide best practices out of the box. They also need to be self-learning, capable of understanding both the context in which they’re being used and how results will impact action.
- Automated. New predictive analytics solutions must be able to handle a large number of decision points without needing to engage resources to manually develop and roll out statistical models.
- Embedded. In order to address the marketer’s need for a scalable decision-making engine, the new wave of predictive analytics solutions must be embedded into marketing processes. These next generation solutions cannot be separate activities that churn out custom results.
Several attempts to create embedded analytics have been tried over the years but typically, these “solutions” were not integrated – leaving marketers to stitch them together in order to get real benefits. Only recently have products with self-learning automated models that are truly embedded and integrated into the marketing process become a reality.
As the industry continues to evolve, expect insights from early inbound marketing products to impact integrated marketing management suites. Then, newly-embedded models will be deployed to help better define outbound marketing audiences, offers and channels. Eventually, marketers will be able to leverage predictive analytics as part of the normal, everyday marketing process. What does the future look like? Here’s an example. Let’s say you take a simple action, such as adding a new digital asset to your system. Shouldn’t your marketing suite be able to do an automated analysis and then pop out a question like, “Did you know that based on our response history, the Digital Asset you checked in could generate a 20% response rate for married customers who have purchased blue shirts in the past six months?”
Doubtless this would make relevant and meaningful sense for your customers. Wouldn’t it also make complete sense for your marketing organization?