In this special guest feature, Dr. Olly Downs of Globys, discusses the missing link between the promise of data science for today’s marketer and actually driving value from real-world applications. Dr. Olly Downs is Chief Scientist/CTO at Globys, a company that has pioneered the delivery of out-of-the-box software that helps companies understand and influence customer behaviors. With 28 US patents, Downs is an expert in big data marketing technology, and specializes in applying abstract analytical ideas from the world of math, physics and statistical science to real-world business problems. Based in Seattle, Olly holds Ph.D. and MA degrees in Applied & Computational Mathematics from Princeton University, and BA, MA and MSci degrees in Experimental & Theoretical Physics from the University of Cambridge, UK.
I don’t want to talk myself out of a job, but to capitalize on the promise of big data, companies need more than a team of data scientists.
We’ve all seen the results in companies like Google and Amazon that are able to provide highly personalized products and services, engage in agile decision making, and quickly generate new sources of revenue.
But today, the average company isn’t seeing those spectacular results. Companies are investing in state-of-the-art data infrastructures. They’ve hired data scientists. But they have yet to see a significant return on their investment.
The missing piece
Most of the big data investment focus to date has been on the underlying infrastructure, while development of the applications that make use of that infrastructure – and that deliver actual business value – has lagged.
Marketing organizations in particular are eager to use big data to drive better results. Today’s marketers want to be data-driven, yet in most cases they don’t have the time or tools to analyze the increasingly large volumes of data that they collect. For this they have rushed to assemble teams of data scientists, with the hope that if they understand each customer and how he or she behaves they can then execute more personalized interactions that impact key metrics, such as revenue, churn rates, consumption, promoter scores, etc.
The value that data scientists deliver in developing sophisticated algorithms that enable intelligent marketing decisions is tremendous.
The reality, however, is that to adopt an entirely new marketing capability – one that leverages modern machine learning techniques to drive always relevant recommendations on a per customer basis – organizations need more than a team of data scientists. What they need is for the work of their data scientists to be productized.
So, why can’t organizations simply invest to build an algorithmic marketing product that makes personalization at scale possible?
For one it’s not easy and requires significant time and investment. At the same time, the problem to solve is not entirely unique which begs the question of whether going at it alone makes sense. Marketers across many industries grapple with harnessing massive amounts of data and traditional personalization, testing and BI tools that are highly manual, and therefore, can’t get to the granularity and scale required to do personalized marketing.
As a result, we’re seeing that companies are looking toward vendors that can provide complete, out-of-the-box solutions – solutions that have been developed based on a pool of R&D efforts and are already proven in market – instead of embarking on the journey alone.
There’s also the challenge of operationalization. Even if organizations hire a team of data scientists, most data scientists aren’t involved in the entire workflow of running personalized marketing campaigns, or worse, they are and the workflow means their job is shuffling “whitelists” of customers around to enable the marketing, and struggling to measure marketing outcomes! This creates the challenge of how to put a new marketing technology in place that is ready to participate in an enterprise infrastructure and drive specific business impact. Data scientists are skilled at working with and researching new algorithms but that’s different than implementing a product that applies and constantly fine tunes those algorithms according to millions of consumers who receive personalized marketing messages every day.
To truly operationalize the work of the data scientists, teams must understand and address all of the processes that surround the application: How do we acquire, store, and manipulate the data? What integrations with existing systems and tools are required to ensure seamless marketing execution and closed-loop feedback on how campaigns are working? How should the user interface look? What training is required? What’s needed for customer service and support? How do we evolve the product to meet future needs?
When it comes together, the results come at lightning speed
Data science can indeed change the world of marketing. With a product that leverages automated machine learning techniques, marketers have the opportunity to gain insight into customer behavior and response that’s not otherwise possible: What makes customers stop buying? What product adjustments help prevent churn? Given past behavior, what is a customer most likely to purchase in the next month? What is the optimal message tone for a customer with certain characteristics? What is the probability that a certain incentive will drive reactivation?
They also have the opportunity to drive remarkably improved performance lift, and as a data scientist myself, that’s what I find exciting. That is, implementing new algorithms in a scientific marketing product and putting that product in the hands of marketers, going from the discovery of unique findings to actually operationalizing a capability that demonstrates real business value, and most of all, seeing first-hand how a strategic asset built on data science enables true competitive advantage.
Sign up for the free insideAI News newsletter.
Maria in addition, the utter confusion created by the several Hadoop vendors add to the chaos. Every one of them have tried to address the simultaneous access need by implementing their own technology on top of Hadoop. Thereby adding vendor lock in and additional complexity.