Leveraging marketing intelligence automata and the human interface

An abridged version of this article first appeared on on October 23, 2015. 

Consumer technologies have led to unpredictable behavior between consumers and the brands that they love. As an engineer who has applied his trade to marketing technology and its automation, I would not be breaking new ground in saying that the last decade has been an incredible ride. The pace of change in marketing is accelerating as the adoption curve of consumer technology steepens. Mobile technology has created an opportunity to reach and influence a consumer at nearly any point throughout his or her day. An increasing number of sensors carried by people and products create a vast amount of data making complex marketing decisions possible. This has led to seismic shifts from mass communication to personalized, individualized and contextualized customer experiences.

Surpassing the wetware of Don Draper 

The traditional modalities represented within companies have been slow to adapt to this new interconnected marketing world. Reliance on software and algorithms is quite different than relying on the wetware of a Don Draper (the fictional chain-smoking executive on the television show Mad Men.) The necessity to have a set of rules that govern each message and interaction with consumers is the antithesis of a world where marketers have phenomenal awareness of consumer motivations and behaviors in a millisecond. Even more importantly, the ability of the wetware of Don Draper to know everything our systems can learn in real-time and recommend the appropriate action is far inferior to the potential of the machine. Our neurons operate at a about 200 Hz. We carry out actions at about 120 m/s or less. We can only store about four to seven things in our working memory at any given moment. In contrast, machines operate in the GHz range, exchange data optically at the speed of light and have orders of magnitude higher working memory capacities. Even more importantly, we evolve through trial and variation. Marketing decisions driven by wetware are slow to evolve and tinkering with the “source code” to iterate into the next variation is much harder than that of software. This has given to the rise of algorithms and marketing decisions driven by the learning machine.

The supremacy of algorithms

The architecture supporting marketing cognition is taking many forms and much work is being done to evolve quickly. I don’t feel qualified to pick a “winner” in terms of analytic techniques or ultimately how “intelligent” various marketing decisions can become. Neither do I really find myself qualified to debate the potential for artificial general intelligence or its applicability. I can give an opinion on some characteristics of marketing intelligence that will prevail in the near term, though.

I believe that we will get much better at building algorithms that can perform recursive self-improvement, dependent on increasing sophistication in discovering emotions, context, data and its structure through consumer sensory inputs. Additionally, we are already seeing a more public and open approach to making algorithms available for commercial use. The outsourcing of some marketing cognition will accelerate the ability to apply proprietary contexts and data in a marketing moment. All of this will push the current limits of marketing decisioning, but not without first working imperfectly. The companies that succeed will be the ones that encourage an environment of trial and error, or more succinctly failure. It would not be failure instrumented by a marketing executive’s strategic decision, but rather the constant iterations of failures and successes happening millions of times over by a machine.

The human interface 

What should we make of the marketer living in this grand melting pot of algorithms and machines? As an engineer, I’ve devoted much to the merits of the machine and automation. Let me balance the scales a bit.

The human interface to cognition architecture is critical to allow us to evolve appropriately. We need to integrate our compassion and inherent ingenuity as a necessary input. While there are many ways the human has to influence the marketing intelligence automata, I’m going to highlight three.

  1. Augmentation of our own cognitive abilities

At least initially all of this must be seen as an augmentation of our own cognitive abilities. It gives us a chance to explore and exploit. This is an extension of the personality of the marketer and your company’s brand.

  1. Enable software visualization of the data

Humans have a highly evolved form of extracting meaning from data that has been difficult to fully replicate through a program in a machine. Primarily this is done through our impressive integration between our brain’s visual cortex and an array of specialized but integrated cognitive functions. We should leverage this advantage by enabling software visualization of the marketing data. There are many software advancements in this area already and the financial barrier to leveraging at least some visualization is nearly non-existent for most companies.

  1. Understand consumer motivation

Marketing technologists should create a “goal ballast” underneath the marketing automata. Take the point of view of the consumer and discover their myriad of motivations. These motivations should become integrated into the goal ballast of the algorithms. Rather than starting from a position of “how much of product can I sell in this moment?”, create discrete algorithmic services to answer “how can I better understand the motivation of this customer at this moment to better service them?” That approach, salted with a bit of random serendipity, will keep the compass pointed in the right direction.

Still, you may ask: What about the ramifications of all of this known and unknown surveillance at the sensory edges? What about the vast amount of personal and proprietary data being moved around? Are we as marketing technologists and market strategists ready for this? What’s my answer? We’re not ready by a long shot. We have to mature quickly.

Implication for technology thought leaders and architects 

I often feel I’m at the vortex of a swirling storm with new data, new data management technologies, new software players, new analytical techniques and new consumer technologies spinning around me.It’s fun if you like change, but no doubt this is challenging. The reality is most companies will delay significant change as long as possible. They’ll deliberate and debate. They’ll form committees. They’ll engage outside consultants that have never built anything like what the company needs. Some will stare at their navel and believe that they will have (or already have) all the answers within their four walls. My simple advice is to just try. Like a sophisticated algorithm, our IT infrastructure is at a time where it needs to go through some trial and error. It’s not a time to be complacent. According to a recent Gartner survey, the most well known mythical panacea for enabling analytics at scale, Hadoop, is actually not accelerating in its adoption. I don’t feel that Hadoop is a panacea, but it does have very real potential for managing data at a scale that allows for insights we wouldn’t have achieved otherwise. Additionally, solutions like Impala, Spark, Cassandra and traditional software like SAS, Oracle and SAP  each have their purpose and reason for existence within the marketing ecosystems we support.

As much attention that is paid to evolving data management techniques, the silence around enablement of open and always adapting marketing algorithms is deafening. There is no sufficient or complete solution to enable an open, service-based, self-adapting marketing intelligence engine that is combined with appropriate policy control enabling the “goal ballast” described above. Much like the SOA approach many of us have taken with the rest of our enterprise software services, we need to create an abstraction of the learning modules our analytics wonks create that hides the complexity that exists in the machine. Right now, it is left to our brightest Data Scientists, Data Architects, Software Architects and Systems Architects to patch a solution together.

In closing 

While algorithms obviously already exist for marketing purposes today (programmatic ad bidding/buying, product recommendation engines in eCommerce, propensity models, sentiment analysis, etc.), what I’m suggesting is that the nature of the way marketers identify and message to a consumer will be less controlled directly by the human and will instead be augmented to a higher degree by machines. There is also a trend that would seem to indicate that the machinations behind the scenes will operate more like a knowledge collective in order to make better and better decisions. There could be wisdom or “unwisdom” in this potential future state. Without the appropriately tailored motivation, the marketing automata could pursue pathways that promote consumption, ignore positive social constructs, marginalize segments of the population or other outcomes that were unintended, but nonetheless harmful to society.

Rather than believe it may all come crashing down in an ignominious collapse, the companies that embrace this change will be the winners. There is a lot for marketing technologists to piece together, but it can be done. If the cognitive architectures are constructed by embedding the appropriate “goal ballast” beneath its use and consistently allow for the human interface to adapt and survey the intentions and outcomes, we will enter a golden age of marketing intelligence that not only benefits companies, but also helps consumers achieve their personal goals.