Erik Severinghaus is the founder and CEO of SimpleRelevance, a Chicago-based company focused on digital marketing personalization. Prior to that he received a patent while in IBM's IT optimization organization, and helped co-found iContact -- a leading email service provider.
Plenty of news stories indicate that 2015 will finally be the year that marketers realize the value of personalization. While this makes sense in theory, my last few years in the industry have led me to disagree with this sentiment. My frustration with marketers' struggle to use data to personalize even led me to begin a company specializing in digital marketing personalization. While this means I'm biased towards advanced personalization practices through algorithmic learning, my experience working at IBM before starting SimpleRelevance underlined the value I found in machine learning. Most sources claim that 2015 is the year that advances in CRM data collection, marketing platform capabilities and multi-channel options will finally converge to improve both the customer experience and marketers' top lines. Although I know improvements will happen, unfortunately 2015 will not be the year when true personalization goes mainstream.
When I say "personalization," I don't mean the ability to put someone's first name into an email or follow someone across the internet with an item from their shopping cart. I'm referring to the ability for marketers to leverage all data at hand to improve the customer experience.
So if personalization works, improves the customer experience and makes marketers more money, then why are so many marketers nascent in how they use their data? There are a few key traps to blame, and a mindset shift that needs to be made before marketers can embrace the changes I see as the future of personalization.
Trap 1: Fear of Change
There's too much data for marketers to process manually, so true personalization requires sophisticated big data processing. And while the largest, most powerful companies -- including Netflix, Target and Walmart -- use algorithms to optimize the way they reach customers, most marketers stick to older methods such as segmenting and targeting. It's not that they aren't capable of big data processing, but rather that they want absolute control and are averse to change.
While following the herd isn't always ideal, in the case of data analytics and personalization, it's a great idea to emulate the major players. For example, during IBM's introduction of Watson Analytics in the fall of 2014, their vice president of marketing emphasized that people were aware of opportunities to leverage their data and analytics, but were hesitant because of the perceived work that comes along with it.
As a result, IBM's Watson Analytics then provided a template to guide users. There's plenty of software and service options available, as well as evidence of machine learning's value; all marketers have to do is research with their needs in mind, and then execute on that charge toward personalization.
Trap 2: Mistaking Personalization for Segmentation
More data means more categories to stick customers into, and subsequently more segments since marketers traditionally add rules and groups in order to personalize messages. However, each new group creates more work.
A perfect example is when a major travel publication created 10 separate email lists for 10 different travel newsletters. By separating their categories, they intended for each newsletter to offer a custom experience. While it may have seemed like a good idea, it multiplied their work by ten.
This publisher eventually decided to use machine learning to better target content to the individual readers, freeing them from maintaining manual lists. The more sophisticated targeting required less manual work, but needed an adjustment from current segmentation practices to let computers do the content targeting rather than targeting manually.
Trap 3: Complicating Personalization
Creating more segments increases work linearly, so optimizing the message at the individual level takes an incredible amount of manual work. Manual segmentation creates more manual labor, which therefore leads to higher costs as human time becomes more expensive.
Personally, I believe that the future is algorithmically driven content personalization. And through my work and the work of others, the tools for that already exist. But before arriving at a machine learning stage, marketers will need to change their mindset by following a few key steps.
To start, marketers should follow the experts' lead. They're experts for a reason, and when others are approaching new technology with trepidation, you want to be the one who is leading the innovation. Second, reframe how you think about data and segmentation: more data doesn't have to mean more segmentation. This is an unending cycle that will keep you from approaching data analytics accurately. Lastly, try not to think about personalization using what I call "the Law of Ancient Personalization."
Trap 4: Following the Law of Ancient Personalization
This is the notion that the complexity of data, amount of work and costs to your company increase linearly with the number of segments you're grouping customers into. This is an outdated way of thinking in part because the number of manual labor hours are limited -- therefore, you cannot substantially increase the amount of content and segments without increasing the money you're spending for the increased labor costs.
The way I see it -- given the newest tools available -- we should begin thinking in terms of the Law of Automated Personalization. This new law says the amount of data, segments and content you can process is not limited by the amount of human time or money you have; it's only limited by the CPU cycles you can throw at the problem. Thanks to Moore's law, those resources are cheap and only getting cheaper. Shifting the targeting system away from human analysis means that infrastructure can continue to leverage the growing quantities of data.
Looking forward, machine learning will be a foundational technology for enabling marketing and IT. Just as the database has become the core fabric of the data center, in five years, personalization technology through machine learning that supplements the database will be just as important. However, this means breaking up with the processes of the past and trusting -- as well as investing in -- algorithmic-based learning.
So if 2015 isn't the year of personalization, then what year will be? It will arrive soon to transform the ecosystem and likely negatively impact those who cling to the status quo. While it may not be happening this year, you should use the time you have now to prepare.