When Talking About Big Data, Small Data Cannot Be Ignored 【Dr. Cheong's Data Insights 9】

Leveraging Social Big Data: The Hotel Industry Can Enhance Reputation and Revenue

After last year's U.S. presidential election, numerous articles emerged online about how Trump's camp used big data to analyze voters' social behaviors, personality traits, and psychological tests. They reportedly used precise targeting and personalized propaganda techniques to help Trump achieve victory. At the same time, traditional polls predicting Hillary Clinton's win failed repeatedly, leading to widespread skepticism and even claims that "polling is dead."

In retrospect, using big data to analyze netizens' behavior and emotions on social networks indeed played a role. Trump's campaign strategy could make quick and flexible adjustments based on netizens' reactions. However, whether big data's role in psychological testing and matching with voter personal data was as simple as rumored remains uncertain. Even the companies providing big data technology cannot definitively explain its effects on Trump's victory.

As for the failure of traditional polls, there have been many post-hoc analyses. The renowned Pew Research Center in the United States provided three explanations: sample bias in polls, meaning certain groups were not covered by the polls. For example, people with lower education levels, income, and political enthusiasm were less likely to participate in telephone polls. When these underrepresented voters turned out in large numbers, it affected the accuracy of poll results. Secondly, due to the political climate at the time, many Trump supporters did not honestly disclose their preference to pollsters. Thirdly, respondents who indicated they would vote but did not actually do so also impacted the accuracy of prediction models.

While traditional polls do indeed encounter issues like those explained by the Pew Research Center, they have still been accurate in predicting election outcomes most of the time. In fact, if we look at Hillary Clinton's final popular vote percentage, she led Trump by 2.1 percentage points (48.1% to 46%), and national polls at the time showed her leading by about 3%. Therefore, the claim that "polling is dead" may not hold water.

Online Big Data, Offline Small Data

Dr. Cheong revisits the above examples to point out that many people today either blindly worship big data and dismiss traditional small data (specifically polling data) or underestimate the role and impact of big data. In this column, Dr. Cheong has highlighted the increasing importance of big data in business, society, and government. Combining big data with small data can create a multiplier effect. The data company hired by Trump also utilized a large amount of personal data when analyzing voter characteristics, such as land registration and car data, shopping data, club memberships, magazine subscriptions, and the churches people attended. Some of this data was recorded, while some was obtained through polling.

So, what is small data? There is no universally accepted definition in academia or industry. To data analysts, consumer behavior data in offline scenarios is small data. To market researchers, data obtained through questionnaires or individual interviews is considered small data. Therefore, small data refers to data collected directly from consumers in real-world scenarios, such as demographics, gender, age, purchase motivations, personal interests, satisfaction, loyalty, values, and brand perception attitudes. The OTAs big data mentioned in the previous article can monitor the overall and individual hotel reputation value in the hotel industry, uncover customer evaluations of hotel services or facilities, and understand customer preferences. This big data can provide the "what" for hotels. If combined with small data from questionnaires, it can explore the "why" and conduct more multidimensional analyses of attitudes and experiences.

In the data era, the combination of big and small data will allow businesses to understand their customers and themselves more comprehensively.

Dr. Angus Cheong Chairman of the Asia-Pacific Internet Research Alliance and Chief Data Consultant of uMax Data

(Originally published in Hong Kong Economic Times, reprinted with permission)