"Big Data × Artificial Intelligence" ~ The Challenge of Quantifying the Atmosphere of Society and Creating New Demand ~ Part 1

Figure 3: Relationship between TV broadcast frequency, search metrics, and social media metrics for the keyword "Goromaru"

 

Exploring the Causal Relationship Between Public Sentiment and Demand Creation

Matsunaga: Television not only influences online behavior but also sparks actions in the market. A clear example: While PM2.5 had been recognized as an air pollution indicator in many regions globally since the late 1990s, it wasn't widely understood. However, in Japan, after TV began extensively covering it starting in early 2013, sales of masks and air purifiers increased, and even the stock prices of companies handling these products rose. Moreover, information that gains public acceptance tends to be broadcast as common knowledge on TV the following year.
I've also analyzed soft drink sales data. Since these products sell better in hot weather, the basic sales strategy is to increase advertising volume in summer to expand market share. When a TV program then broadcasts, "It's hot today, so please stay hydrated to prevent heatstroke," soft drink sales surge beyond what temperature alone would drive. Furthermore, from December to February, TV exposure related to dryness increases across various contexts, such as dry skin, influenza, and winter dehydration. Timing TV commercials to appeal to dryness issues in sync with this prevailing societal mood can even boost purchases during the colder winter months.
However, reading the public mood is extremely difficult, except for top-tier marketers, planners, and creators. By analyzing the public mood, we believe it becomes scientifically possible to hypothesize the timing and messaging of advertising campaigns, thereby improving marketing precision.

Yamamoto: That's right. As a scientific approach, let me share an analysis example using tomatoes. It's often said that products featured on TV sell well in stores. Indeed, there appears to be a positive correlation between the number of TV exposures for tomatoes and online searches/mentions, as well as in-store purchases (Figure ④).
However, this alone doesn't resolve the chicken-and-egg question: is sales driven by TV coverage, or is coverage driven by seasonal demand? To address this, we analyzed individual purchase histories (ID-POS), segmenting consumers into light users and heavy users based on tomato purchase frequency (Figure ⑤). This revealed that during periods of increased TV exposure, the share of light users grew. Notably, this share expanded further in April 2015. The "Functional Food Labeling System," which allows scientifically substantiated claims about food functionality, was featured on TV. This likely spread awareness of the benefits of nutrients like lycopene found in tomatoes, prompting consumers who previously bought few tomatoes to take action. This analysis provides a clear example of causality, demonstrating how television prompted action.

Matsunaga: In this case, the ability to track individual consumers is crucial. It allows us to understand exactly what messages motivated the light consumer segment. While data scientists are often perceived as experts in complex statistical analysis and modeling, I believe the key lies in selecting the right data and determining the appropriate analytical approach.

Figure 4: Comparison of TV Programs About Tomatoes and Search/Sales Metrics

---

Figure ⑤ Comparison of TV Programs About Tomatoes and Purchasing Demographics