"Big Data × Artificial Intelligence" ~Quantifying the Atmosphere of Society and the Challenge of Creating New Demand~ Part 2

Advances in cutting-edge technologies like artificial intelligence (AI) are expected to not only quantify and structure big data like societal trends, but also create new demand beyond mere prediction.
So, what specifically becomes possible when big data and artificial intelligence are combined? How is new demand created?
Satoru Yamamoto, President of Data Artist Inc., involved in system development, and Hisashi Matsunaga, General Manager of the Intelligence Development Department at Dentsu Digital Inc. Marketing Center, will unravel the mechanisms of "deep learning" and the potential of the "demand creation engine," alongside predictions for "products that will hit big in 2016."

 

Predicting Hit Products for 2016

Matsunaga: It's obvious that masks and air purifiers sell when PM2.5 levels rise, and that soft drinks sell when heatstroke cases increase—if you think of it as a one-to-one relationship. However, the atmosphere in society is composed of various keywords, making it difficult to grasp all these relationships. Furthermore, it's not realistic for people to continuously analyze the content broadcast on TV every day.
Therefore, we leveraged AI—Yamamoto's area of expertise—to tackle these challenges. As part of this effort, we predicted products likely to hit big in 2016.

Yamamoto: First, regarding what products might emerge next, we analyzed the fresh fish, fresh meat, and dairy product categories. We focused on cheese, whose TV exposure had been rapidly increasing since around June of the previous year (Figure ⑥). Around this time, Japanese cheese was served to world leaders at a summit held in Germany. News reports highlighted that Japan produces high-quality cheese worthy of global pride. Furthermore, television coverage noted that premium cheeses were gaining attention and were prominently displayed in stores. Based on this, we predicted that potential demand for cheese in the public's mind had increased compared to the previous year.

Figure ⑥ Trend in TV Programs About Cheese

Deep learning to determine when, how, and what to sell

Matsunaga: Up to this point, we're only predicting what is likely to sell well. While cheese sales might be higher than usual due to the heightened latent demand created by TV coverage, our goal is to generate new demand.
Using deep learning, we can suggest when, how, and what to sell, right?

Yamamoto: That's right. Using deep learning (see explanation), the latest AI technology, allows us to automatically quantify and structure the features within data. It's well-known that Google trained deep learning on vast amounts of cat photos, enabling it to automatically recognize features like eyes and whiskers.
This time, we modeled the neural networks of the human brain to capture the nutritional information patterns stored in people's minds, as shaped by television. We then derived the similarity between this neural network and the feature network of cheese.
The results showed that blue cheese and other blue-veined cheeses matched the data from 2015. However, focusing specifically on March, fresh cheese exhibited a high similarity (Figure ⑦).
Fresh cheese contains abundant lactic acid bacteria, which are said to suppress allergic reactions, including hay fever. In March, a nutritional information network containing hay fever was formed in people's minds, and the results of its automatic learning are reflected here.

Matsunaga: While we presented cheese as an example this time, it's also possible to target specific products. By combining the prevailing social atmosphere with deep learning to automatically indicate "when, how, and what to sell," we believe marketing ROI (Return on Investment) can be enhanced.

Figure 7: The "Koto-Mono Engine" that matches societal trends and events ("Koto") with products ("Mono") using deep learning

Challenging the Realization of a "Demand Creation Engine"

Yamamoto: The algorithm introduced today, which derives "when, how, and what to sell" based on societal trends, is also being piloted in Data Artist's conversion maximization tool, "DLPO."
This enables not only LP optimization based on site behavior but also real-time content delivery reflecting the ever-changing atmosphere of the world.

Matsunaga: To maximize conversions, the creativity of the content delivered through LPO, along with the production system and operations, are also crucial. Online, real-time ad delivery to individual audiences is possible, so delivering appropriate content to those who haven't yet taken action can also stimulate that behavior.
In January of this year, Dentsu Inc. established a Digital Marketing Center, bringing together digital talent from across the company to build a framework for delivering integrated digital services. We aim to organically combine digital initiatives to create new demand in response to latent market needs, thereby contributing to our clients' success.

---

Deep Learning
Deep learning is a field of artificial intelligence that mimics the structure of the brain—specifically, the connections between neurons—through neural networks. It enables the automatic quantification and structuring of abstract features within data, something previous AI could not learn. In recent years, improvements in algorithms have dramatically increased the size of the brain that can be simulated, leading to significant progress.