In the episode, “When Knowledge Conquered Fear” of Cosmos, Neil Degrasse Tyson paints a picture of wonderment as early humans gazed up at the expansive, overwhelming, and mysterious night sky. Human intelligence and an aptitude for pattern recognition developed naturally; those who could tell prey from predator and poisonous plant from nutritious were the ones who survived to pass on their genes.
Pattern recognition shaped human history in a powerful way. The ability to recognize patterns in our environment led to our ancestors learning to “read the calendar in the sky.” By identifying patterns, they learned migration timing and trends, weather cycles, and how to navigate by the stars. The cycles of the sky informed people of the cycles on Earth. There was a direct human connection to the sky. People gave constellations names and meaning. Before we knew what comets were, these big flaming balls in the sky were considered warning signs of impending doom. In fact, disaster originates from Ancient Greek: dis- (bad) + aster-(star) as ancient civilizations blamed calamity on the misalignment of the plants. We look for patterns to find meaning in the world around us, but this can lead to false pattern recognition.
Over centuries, we developed tools to aid our ability to measure, record, pass on knowledge, compare, and predict. Curiosity, communication, and sharing are foundational human instincts that allow for progress. Initially, records and data were stored in books and expansive libraries and were difficult to aggregate and use.
Society at large has an insatiable desire to know the future; however, many people still gravitate toward palm readers, horoscopes, astrology, and even fortune cookies. Math and science aren’t always the first choice most people gravitate toward when faced with questions about the future. Perhaps it is seen as boring, too difficult, or that prediction is impossible.
Enter Big Data
The invention of the computer revolutionized our ability to store large amounts of data in a sourceable and useful way. We built machines that can learn from experience. We have been recording what we do know and generating epic amounts of data that is cataloged with regimented discipline. An unfathomable metric of 2.5 quintillion bytes of data is recorded per day! This unnatural resource is driving the future of our society.
In the New York Times article “Big Datas Impact in the World”, MIT professor and economist Erik Brynjolfsson compares the current revolution of data measurement to the revolution in biology triggered by the invention of the microscope. The microscope allowed us to view things previously thought to be too small; with big data, we are seeing things formerly thought to be too large.
The movement to make more data public has also gained momentum, allowing for more collaboration and insights. (To see this in action, check out KDnugget datasets.) People are democratizing data.
Predictive Analytics in Business
In the world of eCommerce, gathering data is only the first step. Now that we have all of this data, what do we do with it? More importantly, how is it actionable? First, we need a way to interpret this data; we find that in predictive analytics.
Predictive analytics helps eCommerce business owners make better decisions. You can probably relate to John Wanamaker, who said “Half my money I spend on advertising is wasted; the trouble is, I don’t know which half.”
Predictive analytics is the practice of extracting information from existing data sets to determine patterns and predict future outcomes and trends. The goal of predictive analytics is to aid in clear decision making. Predictive analytics learns from experience (data) to predict the future behavior of individuals to drive better decisions in a targeted way on the individual level.
Through keen observation, we learn about individual characteristics and find indications about future behavior. Characteristics could be age, gender, email domain, years of school complete, region, etc. A simple linear predictive model weighs each characteristic to determine a score that indicates ideal target individuals. For instance, a male user could increase the score by 35.8, under 35 could increase the score by 12.2, hotmail users could decrease the score by 10.5, and so on. Other approaches include using qualifying if, then statements and decision trees.
Predictive Analytics: Application
Here are some examples of business use of predictive analytics:
- Targeting Direct Marketing
- Customer retention with churn modeling
- Predictive Advertisement Targeting
- Employee Turnover
One of the most memorable examples of Predictive Advertisement Targeting was a model created by Target. Target came up with a clever way to predict which shoppers were expecting parents and marketed relevant products to them based on this profile. Target was able to gain insights from gathering information that seemed innocuous at face value.
Many stores track customer buying behavior in-store and online using credit card numbers, email addresses, surveys, etc. Grocery stores have been creating loyalty programs for discounts and coupons for ages. Targeted coupons based on buying history is nothing new.
Target took this idea one step further by marketing more broadly to expecting parents and not just targeting specific product coupons based on buying history. To gather data about what expecting parents had in common specifically, Target first created a maternity registry. Target realized only a small portion of expecting parents would register, but they found a wealth of data hiding among those who did. They analyzed what these registered shoppers had in common, and then compared those results to the buying behavior of unregistered shoppers and identified who to market to in their new parent-specific advertising. Genius, right?
The campaign was a huge success. However, they didn’t factor in age or add additional qualifiers for ads being sent to expecting parents. A distraught father found out his daughter was pregnant after receiving mail addressed to his daughter with advertisements for maternity clothes and baby items. Many media outlets covered the story and the field of predictive analytics learned an important lesson regarding ethics, privacy, and the power of these new tools and insights.
The Future of Predictive Analytics
As the technology and methods evolve, predictive analytics will only become more accurate. Blending successful predictive models is the future of predictive analytics. Aggregating more in-depth information will become vital to remaining competitive as an eCommerce retailer. The industry term coined for this process is meta learning. Netflix had a dataset contest in which 3 teams of talented modelers banded together to blend their models, creating a superior tool that won them the $1M prize. (So don’t be surprised when your Netflix recommendations become even more personalized!)
It’s more important than ever to understand your customer base and to harness that information for smart marketing. If you’re interested in learning more, I highly recommend starting with Predictive Analytics: The Power to Predict Who Will Click, Buy, Lie, or Die by Eric Siegel.
Finding ways to mine data and make data actionable is one of many benefits to having an agency partner like Gauge. You don’t need to become an expert on datasets, but to be as successful as you can, you do need to know someone who is.