Sign up for our newsletter and get the latest big data news and analysis.

Interview: Globys Answers the Question–Can Big Data Go Too Far?

While data analytics has many highly useful applications, some believe that when not used properly it can have deleterious effects. Globys uses analytics to gain insight into the consumer and offer them well-targeted and relevant marketing campaigns that benefit the user. We caught up with Dr. Olly Downs, Chief Scientist at Globys, to learn more.

insideBIGDATA: Olly, Globys uses data science as a means to enable contextual marketing. Can you tell us what all of this means exactly?

Olly Downs

Olly Downs

Dr. Olly Downs: Our contextual marketing platform, Mobile Occasions, leverages our customers’high-volume behavioral transaction data to gain deep insight into subscriber behavior, the platform then manages, executes and dynamically optimizes marketing campaigns that help maximize long-term revenue and retention of those subscribers, leveraging that insight.

The key elements differentiating our capabilities are;

 

  • DISCOVERY: Automated behavioral analytics and unsupervised clustering that uncovers unique insights
  • EXECUTION: Adaptive always-on experimentation with 1:1 dynamic targeting of context-based offers that trades off exploring for new learnings vs. maximizing the benefits of strongly-performing treatments
  • OPTIMIZATION: Dynamic machine learning on marketing response and behavioral events to maximize performance

 

insideBIGDATA: What is the underlying technology for these marketing solutions?

Dr. Olly Downs: We use Hadoop, Hbase and Impala together with proprietary application code written in Java that enables us to dynamically manage complex combinations of marketing experiments, and proprietary machine learning and data science code written in Python that drives unsupervised customer behavior discovery and closed-loop learning and refinement of campaigns.

insideBIGDATA: This is really useful stuff for organizations big and small it seems. Can data analysis go too far in your opinion?

Dr. Olly Downs: As with all the major technological advances there tend to be applications to the good and the bad.  Our focus is on optimizing experiences to long-terms goals our customers have for our subscribers.  You can’t successfully achieve long-term goals if you annoy customers; interact with them at inopportune times with messages that are out of context and do not resonate.  So I tend to view our application of advanced data sciences as a force for the good.  For example, with the traditional approach, a mobile  operator sends more than 12 marketing messages per subscriber in a 24 hour period from a library of about 1,000 messages – and success is response.  By applying Mobile Occasions, the operator moves to 3-6 interactions per month with a subscriber on a highly individualized basis from a possible 10,000-100,000 messages – and success is the retention of the customer and total revenue over the next 3 months.

Data analysis can go too far when its goals are misaligned with the data being analyzed – for example, the goal being to sell data about individuals to third parties – the data about individuals should be being analyzed to benefit the individual.

insideBIGDATA: Can you possibly give us a real world example of this?

Dr. Olly Downs: I think one of most poignant examples are technologies focused on the discovery of personally identifiable information (PII) from public sources on the internet.  I recall being somewhat horrified by the sources of information companies like Whitepages and Intelius had on me a few years ago, and at the same time realizing how easy it is to find – take for example, full name and address published on my county’s tax parcel GIS site.

At Globys, we never receive PII from our customers about their subscribers, nor do we aggregate data across information sources with the goal of discovering such information.  Further, our business is focused on the success of our customers’ relationships with their subscribers rather than on selling subscriber information to any third party.

insideBIGDATA: If we go too far in over-analyzing data what risks do we take on?

Dr. Olly Downs: The risk we take on is that the data analysis creates a liability rather than an asset.

insideBIGDATA: How do you balance data analytics and human instinct?

Dr. Olly Downs: Great question – this is something we are working very hard on at Globys.  As our Mobile Occasions solution is directed at helping marketers at our customers, its primary goal is to allow the marketer to focus on their strengths – creative message language, offers and incentives and embedding the knowledge and hypotheses they have about their domain (market dynamics, customer base) into their marketing program – while letting the ‘machine’ take care of the elements of scientific marketing – audience behavioral segmentation and discovery, campaign analytics, experimental design and optimization, and performance measurement – that otherwise bog down the creative process.  At the heart of this is making sure the marketer can explore and understand what Mobile Occasions discovers and then learns about their subscribers, so that they can continue to expand and refine their domain expertise and their creative inputs to their marketing program.

The challenge is to get the humans in the loop at the right points in the process, so they feel empowered, deeply informed and in control.

insideBIGDATA: These points are quite important as we learn more and more about data and its uses. What else should we be concerned with as Big Data marches on?

Dr. Olly Downs: I would be looking for the early places where Big Data truly delivers ROI – it will be the end-to-end use cases where the object of the data and the goal of the data analysis are aligned where there will be a virtuous cycle.  Until those use cases emerge I think we’ll continue to see a very conservative approach to Big Data technologies in the mainstream where the business benefits tend to focus on reducing costs.

Resource Links: