Examining Big Data's Potential In Predictive Marketing

The cost to store and process large amounts of data continues to fall. At the same time, new, simple software tools are emerging that allow companies to meet the single achievement of all great marketing campaigns: accurate prediction and high ROI. How do these new software tools improve marketing? This blog post gives a cursory look, and our new white paper, Examining Big Data’s Potential in Predictive Marketing, walks you through essential details.

Marketing-specific applications for big data include improved prospect identification, simultaneous ad campaign tests, and highly personalized message creation. With large-scale data analysis tools becoming more accessible, business analysts are predicting continued growth for big data. Tech guru Mary Meeker expects the amount of content and data on the Internet to double by 2014; media firm SiliconANGLE predicts that in 2013, 50 percent of businesses will make use of external data – data that companies purchase, rather than generate in-house – to make business decisions and drive marketing programs.

A Test Case in Big Data

Large data collections hold tremendous potential for marketers when combined with other data, such as internal customer information.

Take this example, based on Microsoft’s fictitious business case, Adventure Works Cycles (AWC): Robert Smith, chief marketing officer at AWC, a bicycle manufacturer, runs an Internet-based promotion offering 50 percent off bikes. The campaign is less than successful and, after reviewing the data related to the ad and subsequent online purchases, Smith realizes the ROI on the ad was low.

Smith then uses Microsoft HDInsight to import AWC’s unstructured ecommerce transaction data, and he compares that to data describing spikes in traffic on AWC’s website. Then, he uses PowerPivot, to compare those spikes in traffic with weather data obtained through the Windows Azure Marketplace. The analysis shows a strong correlation between warm, sunny weather and web purchases.

For more information about Smith’s analysis and the details he used to make decisions for his campaign, download our white paper.

Defining Big Data

RDA defines big data as follows: a set of data with at least one of the following characteristics – high volume, high velocity, high variety, or high value.

HIGH VOLUME means data that can’t be economically stored in a traditional on-premise database because the cost of storing it is greater than the value it currently provides.

HIGH VELOCITY refers to data that requires a high processing rate but is not large enough to classify it as high volume. For example, an energy company might have devices that include sensors. Each sensor sends a small amount of data every few seconds that is processed by the energy company in order to optimize the device and maximize power production during times of high load.

HIGH VARIETY refers to unstructured data, such as emails or Twitter posts. The make-up of this type of data doesn’t lend itself to traditional relational storage and analysis, but semantic mining can be successfully performed on it with the right software tools.

HIGH VALUE signifies the ability to harness information in novel ways to produce useful insights of significant value that were not previously available.

Data Sources

The big data that companies use for marketing purposes can come from inside the company, another website, or a marketplace. The following examples give you an idea of potential sources of data for your next marketing campaign:

INTERNAL

Web application log data
Customer email
Customer feedback (from a form)
Customer feedback (from call center audio recordings)
Corporate documentation
Contracts
Status reports
Trouble tickets
Data from a product that includes a sensor

EXTERNAL

Websites

Utility providers (gas or electric)
Zillow
News feeds
Facebook
Twitter

Marketplaces and Repositories

Windows Azure Marketplace
Amazon Web Services
Databib
DataCite
Figshare
Linked Data
Reddit
Quandl

It’s All About Better ROI

In our hypothetical Adventure Works scenario, Robert Smith ultimately purchases data from Weather Trends International, which he finds in the Windows Azure Marketplace. The data includes historical daily maximum temperatures, minimum temperatures, precipitation amounts, dew points, sea level pressures, wind speeds, and wind gusts for thousands of locations around the world – all of which he feeds into his new marketing campaign. That, in turn, leads to a more profitable campaign.

To learn more about making big data work for you, download our white paper, Examining Big Data’s Potential in Predictive Marketing.

image: predictive marketing/shutterstock