How Can Thick Data Improve the Way We Use Big Data?

Big data can help us solve problems, optimize services, predict the future and learn more about ourselves and human behavior. But when big data cannot answer a question, or makes an incorrect prediction, what if it is because data is not telling us the whole story?

In a Ted Talk and an article on Medium, Tricia Wang, a technology ethnographer, implores us to realize that, even in the age of data-driven decision making, big data can have some serious blind spots, and often suffers from a loss of context through valuable variables and trends that defy quantification.

Insights on contained systems, such as electricity grids or delivery logistics, have established big data’s reputation for success. Although this reputation has attracted investment from a huge array of companies, Wang claims, 73% of big data projects are not profitable, because big data projects frequently do not produce the insights that they seem to promise. For dynamic systems, especially those that depend on human emotions, behavior, and how people are impacted and interact with their environment, certain variables are difficult to measure and model.

She describes how opportunities can be gained and losses diminished by using big data in conjugation with “thick data”, observations that use qualitative methods to understand human emotions, behaviors and interactions with context. For example, the realization by ethnographer Graham McCracken that Netflix users actually enjoy binge-watching, led Netflix to redesign their user experience, improve their business and transform the way we consume media. Through qualitative research and collecting stories about how users interact with their product, Netflix was able to go to their data science team, validate these findings, and act on this competitive insight.

Wang summarizes that while “Thick Data loses scale…Big Data loses resolution.” While big data analytics can have a huge sample size, recognize large-scale patterns, and benefit from the strengths of machine learning, thick data uses small samples sizes, but understands subjects at a much greater depth, and benefits from aspects of human intelligence and pattern recognition that still elude computer programs. As data scientists, the more we understand the limitations of our methods and pay attention to thick data, the more we can recognize the factors and variables our data is missing. With a more holistic perspective, data analysis is better poised to uncover more truthful and powerful insights.