ibank

With the emerge of the age of digital marketing, big data marketing is becoming BIG. But in the end, the algorithm codes of the big data are still set by humans, and therefore human error and bias are inevitable. To get the right insight out of all that data mess, it is important to realize that the ultimate obstacle is human.

George Orwell's 1984 Big Brother is no longer a fiction.

Someone is always collecting, storing and using my data, even without my knowledge. Creepy.

To make matters worse, what if you would find out that the information is used illegally?

The buzzword of the digital marketing era undoubtedly is 'Big Data.'

Until now Psychology dominated marketing strategies. But with the expansion of the smartphone, the dream of handling a tremendous amount of data has become real. Therefore strategies based on 'big data' gain more credibility than the traditional psychological principles. Out of the big mishmash of data, only the needed information like location, purchase pattern, etc. is sorted out, collected and analyzed to predict the future.

Of course, negative opinions also exist. For example, Cathy O'Neil, a professor at Harvard University and a prominent data scientist, warns that the misuse of data is as lethal as a WMDs and called it 'Weapons of Math Destruction' in her book. Yuval Harari sarcastically mocks the rise of the data religion in his book 'Sapiens.'

The problem lies in the 'algorithm' that is used to analyze the data. People tend to believe that artificial intelligence, machine learning, etc. is neutral, objective, accurate, and scientific. So it is easy concluded that the conclusion AI draws is superior. But it is important not to forget that it was a human who programmed the code of this algorithm in the first place.

Machines replaced the hands and feet of humans through the first, second, and third industrial revolution. We are now standing at the branch of the fourth industrial revolution which tries to replace the human brain with machines. Therefore, process, human prejudice, stereotypes, misunderstandings, and bias can be reflected in the data selection process. As long as you use an incorrect algorithm, you can't but get wrong results, regardless of how good the quality of your data is.

The best-known examples of a false algorithm are the New York Police 's Criminal Predictive Model and Google' s Flu Trends.

First, the example of New York Police. Based on crime statistics data, the NYPD predicting criminals and concentrated police forces in areas where crime was expected most likely to occur. Unfortunately, this algorithm caught only misdemeanors, which is the majority of crimes committed, while financial crimes slipped through. More homeless beggars, drinking minors, and junkies were found, and the area became crime-prone. Leading to even more police force patrolling the area and finding out more misdemeanors. For example, Afro-Americans and Latinos aged 14 to 24 account for only 4.7% of New York's total population. However, 40.6 percent of the testimonies that were inspected by the NYPD police were Afro-American or Latin, and 90% of them were innocent. Cathy O'Neil strongly criticized that the "negative feedback loop" is working because of the wrong algorithm, "the police activity itself produces new data, and this data justify more police activities again."

The next example is the epic failure of Google Flu Trends.  Google used 'Google Trends' to estimate influenza activities in the United States. It attempted to make accurate predictions about flu activity, and the essential idea was published in Nature. And then, GFT failed—and failed spectacularly—missing at the peak of the 2013 flu season by 140 percent. One of the biggest reasons for the miscalculation was the "flu emergency" issued by the US in 2012. Even unaffected people started flu-related searches leading to a  misinterpretation. Google Trends misinterprets this number and overestimates the rate of colds. The poster child of big data into the poster child of the foibles of big data.

These two examples are showing the limits of Big Data. Both cases are also a lesson for using Big Data as a marketing strategy. The ultimate obstacle to Big Data is the human being. Never forget the customers hidden beyond quantitative data and that it is also a person that analyzes and utilizes the data.

Poking around in the wrong direction is as meaningless as poking the right place with the wrong tool. Empty communication is as good as a spam message, which is, in other words, nothing but a total waste of budget.

That's why it's important to create insight out of the enormous amount of data, set the right goals for algorithms, and audit the entire flow.

Here are five essential principals for understanding data at work from Thomas H. Davenport, Professor at Harvard Business School.

- Identifying and Framing the Analytical Problem

- Working with Quantitative People

- Understanding Different Types of Data and Their Implications

- Understanding Different Types of Analytics and Their Implications

- Exploring Internal and External Uses of Analytics

* Sources

- Cathy O’Neil,

- Thomas H. Davenport,

- Dublineryh의 포스팅 <빅데이터는 실패하였다? 빅데이터의 가능성과 한계>