Big data is everywhere, both literally and figuratively. On the figurative side, it’s hard to escape the hype about big data, the term appearing in headlines of major national newspapers and business publications (most recently the cover of the "Harvard Business Review"). But the hype is happening for a reason, and that’s the amount of data we’re all collectively generating in our off- and on-line worlds.
Every two days we generate as much data as 350,000 times the entire printed catalogue in the U.S. Library of Congress. That includes data generated by 4.8 trillion online ad impressions in 2011 alone, and the 294 billion emails sent every day. All of this data stands to provide incredibly valuable insight about the habits and intentions of consumers, if it can be interpreted. This ultimately is the promise of big data, the ability to capture, combine and gain insight from massive amounts of structured and unstructured data.
Big Data, What’s it Good For?
First things first, what exactly is big data good for? The answer is the same as most forms of marketing intelligence: better insight about customers, the ability to more precisely segment customers into meaningful groups, and target offers with a higher degree of response.
Big data is particularly relevant for predictive analytics where the goal is to model the intention or propensity of a buyer to purchase. In fact, recent research (Go Big or Go Home? Maximizing the Value of Analytics and Big Data) found that 90 percent of the organizations wrestling with big data were also using predictive analytics. Part of the reason for that is that big data is just so big it’s no longer possible to spot trends and patterns using the simple human eyeball.
Something else is needed and that something is predictive analytics. This can inform everything from which product or service to offer a prospective customer in a campaign, to when to send the next email and what the subject line should be. Aberdeen’s predictive marketing analytics study Divide & Conquer: Using Predictive Analytics to Segment, Target and Optimize Marketing (December 2011) found that improving targeting of marketing offers and obtaining a 360 degree view of the customer were the top two strategies for marketers using predictive analytics.
Two significant themes emerge from Aberdeen’s research on marketing data analytics that are helpful in considering the big data trend for Marketing.
- Top-performing companies (i.e., “Leaders”; see sidebar) are more likely to incorporate both structured and unstructured data in their predictive models
- Leaders are more likely to give access to business users, speeding time-to-deployment of analytics models and opening new applications for analytics
Fueling the Database of Intentions
Data is the fuel of analytics; as the CRM manager for a European utility company put it, “To a large extent, the quality of the model is only as good as the quality of the data that you have.” As such, one would expect data-oriented capabilities to be widely adopted among users of predictive analytics, and indeed they are.
As the figure below shows, Leaders are more likely to have access to a wider range of data than their more poorly-performing peers in several categories. Several things to note here. The first is the access to both transactional and behavioral data, which is particularly important in a marketing context as it can provide actionable insight even when the buyer isn’t known, as well as support real-time application of analytics, such as online ad display or product offers. The second is the access to unstructured data both internally (such as from a call center) and externally (i.e., from social media).
Source: Aberdeen Group, December 2011
The second trend shaping the future of big data for marketing is ease of use and ease of integration. The benefits here are the ability to more quickly gain actionable insight and then to operationalize such data-driven insight in campaign tools.
Nearly half of Leaders (46%) in Aberdeen’s research found their predictive analytics solutions easy to integrate with other applications or business processes, while only a quarter (26%) of Followers enjoyed that same ease of integration.
Integration is crucial as predictive technologies are rarely used in isolation. To be effective, the output from predictive models is usually injected into business processes. For example in marketing, predictive applications can be used in real-time to determine the next best offer, either on the web or in a call center environment. Similarly, for a direct mail campaign, the output from predictive modeling will be fed into the mailing solution.
Overall, forty-five percent (45%) of study respondents indicate that a lack of key technical and/or mathematical skills presents a significant barrier to the use of predictive analytics. When predictive modeling skills are so scarce, one remedy is to place as much of the predictive modeling task as possible into the hands of business managers.
Aberdeen's research found that Leaders are twice as likely as Followers (43% vs. 21%) to have data mining tools that can be used directly by business managers without help from statistical experts. If the tool's interface hides the underlying algorithms from the user, then staff with little or no knowledge of statistical analysis can engage directly with predictive models. In this way, enabling business users with marketing expertise, but little or no knowledge of statistics, to manipulate models directly can improve marketing productivity. Without this hands-on approach, marketers are reliant on the rare skills of dedicated predictive modelers and statisticians to help them target marketing campaigns and assess their performance.
Providing solution templates and examples can also help marketers work directly with predictive models and speed the path to results. Templates serve as building blocks. As such, templates provide a springboard that helps marketing applications get completed faster by people that lack in-depth technical knowledge and skills.
Big data is all the rage, and marketers are just beginning to both adapt existing data analytics practices and models to incorporate high volume, unstructured data in real-time, while unlocking increased business value through better ease of use and ease of integration.
The traditionally high cost of business analytics -- both in terms of technology and talent -- has limited their application to high-value, high-volume use cases. Increased ease-of-use and decreased solution costs are thus not only making marketing analytics relevant for a wider audience of users, but a broader set of use cases as well.
Editor's Note: Our focus on turning data into marketing action is coming to a close. Interested in reading more?