Is Big Data Dead?!
Is big data not as useful as we think in solving Business Analyst problems? The writer Martin Lindstrom, through his book “Small Data,” suggests that big data might not always be the key to unlocking answers to business questions. According to him, big data can sometimes be uninformative or even misleading. Instead, Lindstrom highlights the crucial role of small data in the business world, illustrating through his book how small data can be effectively utilized to find solutions to real-world business problems. This perspective not only questions the existing reliance on big data but also underscores the importance of small data in gaining meaningful insights and making informed decisions in the business domain.
I found the “Small Data” book very interesting and want to analyze whether big data is not informative or even misleading.
1. Small Data Explained
Small data is primarily about the nuanced observations and insights gathered from the environment around us, as highlighted by Martin Lindstrom in his book. One example is his analysis of fridge magnets in Russia and Saudi Arabia, and how these seemingly trivial items can give significant business insights. Additionally, Lindstrom often draws upon experimental findings related to human behavior, such as the “Counterclockwise Test” by Harvard psychology professor Ellen Langer. These experiments, common in psychology and behavioral economics, offer deep insights into human behavior without relying solely on surveys.
Lindstrom humorously notes his lack of formal training in social science, psychology, or detective work, yet acknowledges that his approach is likened to that of a forensic investigator of small data. This investigative stance stems from a lifelong habit of keen observation and analysis, beginning in his childhood. His book is a testament to the power of personal experience and analysis in unlocking business solutions.
However, it’s important to recognize that Lindstrom’s methods, deeply personal and derived from his unique experiences, might not be easily replicable in other business contexts. His reliance on small data often involves a very individual approach to understanding people and situations, which could pose challenges when attempting to generalize these methods to solve different business problems.
Distinguishing Experiment from Small Data
Collecting data through experiments is a common method in behavioral economics and psychology. The main challenge with this approach is ensuring that the experiment is conducted correctly and that participants are honest about their behavior. This type of experiment is utilized to analyze people’s responses, aiming to understand the correlation and causation behind their actions. While I do not intend to delve into the methods of data collection and analysis within an experiment, it’s important to note that researchers are exceedingly cautious with the methods and tools used to experiment. This is because a minor error or incorrect treatment could lead to inaccurate conclusions. Although Martin Lindstrom is not a psychologist and is not bound to prove the correctness of his analytical methods, he simply observes people and performs his correlation or causation analysis based on these observations. Lindstrom’s analysis of people’s sexual behavior and the direction of their toothbrushes is intriguing. I hope that some researchers will conduct experiments to determine if there is indeed a correlation between these factors. Such findings could be beneficial for some individuals. This observation is Martin Lindstrom’s small data, but if someday researchers can prove the correlation, then it will be experimental data.
The Role of Biases and Heuristics
Get Huseyn Abdullayev’s stories in your inbox
Join Medium for free to get updates from this writer.
In this book, he also referred to some experiment results that revealed biases. Biases are phenomena where people do not behave rationally but rather exhibit some irrational behavior, even if this behavior causes them to lose benefits. Another type of irrational behavior is heuristics. Heuristics describe the phenomenon that people show different behaviors in different situations. For example, the book mentions the “Counterclockwise experiment” by Harvard psychology professor Ellen Langer, which demonstrates how people’s behavior changes when they are placed in a different environment. Cognitive biases and heuristics are crucial topics for analyzing behavior in economics and business. If you’re interested in biases, the Hidden Brain podcast is a great resource to explore.
Correlation vs. Causation: A Critical Analysis
Another important point I think we have to talk about is confirmation bias. Confirmation biases occur when analysts try to find evidence and information in their data that proves their initial thought. It means that we focus on the parts of the data that support our first thought. Therefore, the problem of not seeing important insights or not correctly analyzing the case is not related to Big Data but rather to the interpretation of the data itself. For example, in the book, the writer states, “Every big data study Lego commissioned drew the same conclusions: Future generations would lose interest in LEGO.” This conclusion is an interpretation of Big Data, meaning the analysts think that due to how impatient, impulsive, and fidgety millennials are, Lego should begin manufacturing bigger bricks. This is a classic example of correlation-causation misanalysis; there could be a correlation between the behavior of the children and their demand for the type of Lego. But the conclusion is that if the children are becoming more impulsive, therefore, we must produce bigger bricks is a misinterpretation.
Such misinterpretations are not inherent flaws of big data but rather result from erroneous analysis. It’s crucial to differentiate between seeing patterns in data (correlation) and inferring a direct cause-and-effect relationship (causation). For example, noticing that people wear less clothing as temperatures rise throughout the year shows a correlation between temperature and clothing choice. However, assuming every correlation implies causation can lead to misleading conclusions, as some correlations may be coincidental, exhibiting either a strong negative or positive relationship. This kind of correlation is called Spurious Correlation. You can a lot of information about it on the website https://www.tylervigen.com/spurious/discover. For example, I put one sample below to show the Spurious correlation between the search for who is Donald Trump in Google and Air pollution in Keene.
Source: https://www.tylervigen.com/spurious/discover
Even the causation analysis is more challenging than correlation analysis. Let’s go back to our example; it is obvious that the weather temperature causes changes in people’s clothing behavior. But imagine that we do not have any prior information, and then determining the direction of causation becomes quite challenging. This means we would not know whether changing temperatures cause changes in people’s wearing style (more or fewer amounts of clothes) or if people’s wearing style influence the temperature to increase or decrease. This kind of problem is very common in business, and big data itself does not provide direct insights into causation analysis. As analysts, we conclude by interpreting this data. Therefore, I think it is not fair to blame big data for misinterpretation of the data.
The Issue of Selection Bias
I think that one of the key points to discuss regarding the book is selection bias. Selection bias mainly occurs due to not applying the correct randomization method while gathering information. In its simplest form, we focus on some parts of the information and ignore the rest. In the book, I think we read about cases where the writer was successful. Of course, his analysis in some cases is very precise and deep, but I am sure that there are some cases where he failed. Therefore, we must be very cautious about the fact that we see only the successful part of his job. Maybe there are some cases in which his analyses were not correct .
Conclusion: A Balanced Data Diet
“Small Data” serves as a timely reminder of the richness and value that small, detailed observations bring to the table in a world increasingly dominated by big data. Lindstrom’s exploration encourages a more balanced approach to data analysis, where small data’s nuanced insights complement big data’s breadth. While analyzing our business problem and gathering the small data we should not ignore the problem that could occur and be suspicious of your conclusion.
