Originally published: June 18, 2014
In the world of big data analytics, there is much talk about the data scientist talent shortage. The implicit message in this discussion is that the analytics problems that businesses, government agencies, private institutions, and academic researchers are trying to solve are inherently science problems. In particular, they are data science problems. Consequently, one needs data scientists to work on the solutions to the problems that big data are now bringing into our enterprises.
One of the primary areas where we see greater use cases for big data and analytics is in marketing. The world of marketing is rapidly being morphed into digital marketing – blanketed by the use of digital data, digital signals, digital responses, and digital campaigns. Marketing analytics (big data science) therefore becomes the tool of choice in addressing five key customer challenges: awareness, acquisition, engagement, conversion, and retention. If one looks at these challenges as a science problem, then the value and power of customer data become evident.
Science of any kind includes the fundamental steps in the scientific method (which is actually a continuous cycle of refinement and improvement). These steps are: data collection, hypothesis generation, experimental design, measurement, evaluation and error analysis, and hypothesis refinement. In the context of business, the use of customer data drives the scientific cycle in marketing analytics – generating and testing models of what attracts customers, what engages customers, what leads customers to conversion, what builds customer loyalty, and what retains customers.
At the June 2014 Useful Business Analytics Summit, various speakers emphasized the importance of an “always-on” experimental (A/B testing) data-driven culture. The science of customer data therefore requires a data science approach: defining objectives for your big data initiatives, generating testable hypotheses that aim to address the above 5 key customer challenges, designing experiments (which probably should include significant A/B testing) to test your hypotheses, analysis and evaluation of results, and model improvement/refinement for further optimization of your marketing analytics strategy.
The corresponding portfolio of marketing analytics (big data science) activities can be as simple as descriptive analytics (i.e., what have my customers purchased in the past?), or as modern as predictive analytics (i.e., what will my customers purchase?), or as advanced as prescriptive analytics (i.e., how can I drive customer purchases that will help my business to achieve greater customer conversions, loyalty, and retention), or as bleeding-edge as cognitive analytics (i.e., how can machine learning be used to create an automated environment – such as that offered by Syntasa – that adaptively responds to customer data in the context of different campaigns, seasons, and business objectives?).
The central defining element of this data science approach to digital marketing is the data – the customer data. What you decide to measure, and how you respond to it, is what will ultimately decide the success or failure of your digital marketing analytics initiatives. Taking a scientific approach (testing, failure evaluation, improvement, and optimization) means that you will experiment with different inputs (customer signals) and different outputs (marketing campaigns), driven by different models (data science components in the automated environment pipeline).
So, before you begin your next marketing initiative and begin designing new campaigns, take a page out of a science textbook: collect data, infer a hypothesis, test your hypothesis (through experimental design and measurement), refine your hypothesis, and optimize your model to maximize the desired outcomes. Marketing analytics and the science of customer data will not let you down. Syntasa can help you get there. The data-driven experimental approach is objective, justifiable, reproducible, measurable, and improvable – by design!
For more from Kirk Borne, follow him on Twitter: @kirkdborne