Keywords

1 Introduction

Chocolate has been increasing in Japan domestic consumption. It is eaten regardless of the age from the adult to the child. It is sold at supermarkets and convenience stores. People easy to get. It has a wide range of prices and there are many variations.

Therefore, it is one of the simple and familiar sweets for the Japanese. We often watch commercials for confectionery companies that selling chocolates. Their companies are advertising TV as one of public relations activities. But, in Japan people who watch television are getting less. According to the Ministry of Internal Affairs and Communications data of 2013, the average viewing time of real-time TV on weekdays was 168.3 min [1]. The weekday average for the last year was 184.7 min. It was 16.4 min (about 9%) decrease. The average for 2011 (including holiday data) was 228.0 min. It was found to be 59.7 min difference.

We are watching over average value on weekdays only in 60’s. The holidays are the 60s and 50s. Teenagers and twenties are the most unviewed. On weekdays it was about 40 to 70 min shorter than the average and holidays 60 to 90 min shorter. According to the data surveyed by the Risky Brand, the TV average viewing time on weekdays was almost flat in the 30 to 49 years old and 50 to 64 years old generation between 2008 and 2013 [2]. The middle-aged and older age is flat from two data. It became that young people do not watch TV. The total advertising expenditure in Japan totaled 5,769.6 billion yen in 2011. Terrestrial broadcasting usage is 1,723.7 billion yen (about 30.2%) [3]. Total in 2013 is 5,976.2 billion yen. Terrestrial broadcasting usage is 1,719.1 billion yen (about 30.0%). The total of advertising expenditure increased slightly in two years. It does not change that it occupies much of advertisement expenditure as present condition. We can think that Japanese television is still the main information dissemination medium. Therefore, even if the TV viewing went down, we clarify whether there is a relationship with purchasing. We revealed it by decision tree analysis.

2 Usage Data

We used attribute data, purchase process data and TV viewing survey data. Their period is from September to October 2011 and from September to October 2013. We used to factor that gender, age, married or unmarried form attribute data. We also created and used the age-based variables. We used questionnaire data on purchase status of September and October regarding chocolate products (Meiji milk chocolate, Lotte milk chocolate, Morinaga confectionery dozen, Meiji almond chocolate, Meiji macadamia nut chocolate) common to both years from the purchasing process data.

TV audience survey data is for 8 weeks from August 27, 2011, to October 22, 2011, and for 8 weeks from August 31, 2013, to October 30, 2013. There are seven main channels in the Kanto area (Tokyo, Kanagawa, Chiba, Saitama, Tochigi, Gunma and Ibaraki) where data was collected. But two of them are public broadcasting. On that account, we excluded them from the analysis. Then, we squeezed from 5 pm to 11 pm and analyzed. Because it is regardless of the day of the week that there are time zones that are seen frequently [1]. In addition, the viewing data include “CM contact data”. “CM contact data” is the number of CM contacts within each period for each sample calculated based on the commercial data for all programs. Even if there are CM advertisements more than once in the same program, it counts as one. CM contact data of chocolate products (Lotte milk chocolate, Morinaga confectionery dozen, Meiji almond · macadamia nut chocolate) whose data are taken in both years are also used.

3 Data Summary

First, we added up each data as the basic statistics from the attribute data. The total number, gender ratio, age, and married or unmarried are as follows.

The data of September to October 2011, the total number is 3109. Gender ratios are 51.1% for males and 48.9% for females. There is almost the same number. The age composition is 1.2% for teens, 20.4% for 20s, 29.7% for 30s, 21.6% for 40s, 24.9% for 50s and 2.3% for 60s. 60.3% of respondents were married, 34.1% were unmarried, and others were 5.7% (Fig. 2). On the second data of September to October 2013, the total number is 3146. Gender ratios are 50.7% for males and 49.3% for females. There is almost the same number. The age composition is 0.7% for teens, 19.7% for 20s, 27.7% for 30s, 25.2% for 40s, 21.7% for 50s and 5.1% for 60s. 57.2% of respondents were married, 37.0% were unmarried, and others were 5.8% (Fig. 3).

Regarding residential areas of respondents, seven prefectures in the Kanto region including Tokyo, the capital area of Japan. With respect to the age, teenagers and 60s have extremely few data compared to other ages. We can not analyze sufficiently. In this report, we analyzed without using data of teens and 60s. Thus, the number of data from September to October 2011 is 3000, and from September to October 2013 is 2964. In each item, it is possible to compare the two data because there is no big difference or bias between them.

From the TV viewing survey data, the total number of programs by day of the week is Table 1. Usually, in Japan, Monday to Friday are weekdays and Saturdays and Sundays are holidays. Both years have many programs on holiday. Regarding the classification of the program, we referred to the “category” of the TV program recording program [5, 6]. The total of program categories is shown in Table 2. Because there was a program with multiple categories, the total number of programs by day of the week and each category total are not equal. Also, the analysis uses six categories above the double line. The reason is that there are more than a certain number of watching. We explain the program category. “Variety Shows” is a show program such as songs, skits, dances, narratives [7]. “Documentary” is made based on actual record without using fiction [8]. “TV Gossip Show” is Information entertainment program including sensational incident. It consists of several corners; the moderator speaks to the guest and progress the program [7]. “Sports” are broadcast programs related to sports such as a game relay. “Hobbies/Education” is a broadcast for school education or social education. And those that directly aim at the improvement of general public education [9].

Table 1. Number of programs by day of week and total
Full size table
Table 2. Number of programs in each category
Full size table

4 Analysis Overview

We explain the analysis procedure of this paper. First, we gave the category information to the program data. Then we created a dummy variable for the day and time on which the program was broadcasting. For example, for programs that were broadcasting from 17 o’clock to 19 o’clock, we set 1 for variable names “17 o’clock” and “18 o’clock”. Next, program data was combined with viewing data. And we compiled it. Furthermore, we made it into a variable which the viewing frequency within the period. For category data and broadcast time data were created variables like Table 3. As for the data on the days of the week were cleared variable like Table 4.

Table 3. Variable and interpretation of variables by frequency (category, broadcast time data)
Full size table
Table 4. Variable and interpretation of variables by frequency (day data)
Full size table

In the questionnaire data, we created the purchasing process data again as shown in Table 5. For those who purchased Meiji Milk Chocolate, Lotte Ghana Milk Chocolate or Morinaga Darth in September or October, those who did not purchase 1 or 1° were 0.

Table 5. Reprinting questionnaire data on purchase actual condition
Full size table

5 Analysis Result

The conditions of the analysis are shown in Table 6. We chose “bought it” in the category of the dependent variable. In all the analysis results done this time, since the P value was less than 0.05 at all nodes, we think the model fit is good. The numbers below the nodes in Fig. 6-1 to Fig. 6-12 are index values. If it is 100%, it means that the purchase probability is 1.00 times compared with the data set. For example, if it is 120%, 1.2 times. The rectangle of each node represents the ratio between purchaser (dark gray) and non-purchaser (light gray).

Table 6. Common conditions of analysis
Full size table
Fig. 1.
figure 1

Chocolate consumption trends

Full size image
Fig. 2.
figure 2

Gender ratio

Full size image
Fig. 3.
figure 3

Age ratio

Full size image

5.1 When the First Independent Variable Is Set to “NEWS”

The highest percentage of purchase in 2011 results node 9. It is a person who watches “NEWS” 2 to 3 times a week and “Sports” once to twice times a week and watches “TV drama” 5 to 6 times a week or hardly watch. Their index value is 134.0%. Node 11 is the next highest percentage that is 119.8%. It is a person who watches “NEWS” 0 to 2 times a week and married 20’s, 40’s and 50’s. Node 1 is the third highest percentage that is 116.7%. It is a person who watches “NEWS” 4 times or more. Node 13 is the worst percentage that is 55.7%. It is a person who watches “NEWS” 0 to 2 times a week and unmarried 30’s.

The highest percentage of purchase in 2013 results node 9. It is a person who watches “NEWS” 0 to 2 times or 7 or more times a week, “Variety Show” 0 to 1 time or 3 or more times a week and married. Their index value is 111.3%. Node 11 is the next highest percentage that is 105.5%. It is a person who doesn’t almost watch “NEWS” but who watch 18 o’clock program twice a week or more and 17 o’clock once or twice a week. Node 8 is the third highest percentage that is 102.3%. It is a person who watches “NEWS” 0 to 1 time or 3 times more and married. Node 4 is the worst percentage that is 83.2%. It is a person who doesn’t almost watch “NEWS” and “Variety Show”.

5.2 When the First Independent Variable Is Set to “Documentary”

The highest percentage of purchase in 2011 results node 8. It is a person who watches “Documentary” 4 times a week or less and married 50’s. Their index value is 122.5%. Node 5 is the next highest percentage that is 121.7%. It is a person who watches “Documentary” 3 or more times and “Animation” once time a week. Node 10 is the worst percentage that is 65.0%. It is a person who watches “Documentary” 4 times a week or less and unmarried 30’s.

The highest percentage of purchase in 2013 results node 13. It is a person who watches “Documentary” once to twice times a week, 18 o’clock program 3 to 6 times a week and unmarried. Their index value is 112.0%. Node 11 is the next highest percentage that is 109.0%. It is a person who watches “Documentary” 3 times a week, “Variety Show” once to twice times or 4 or more times and married. Node 10 is the third highest percentage that is 101.3%. It is a person who watches “Documentary” 2 or more times a week, “Variety Show” once to twice times or 4 or more times and unmarried. Node 6 is the worst percentage that is 84.1%. It is a person who doesn’t almost watch “Documentary” and unmarried.

5.3 When the First Independent Variable Is Set to “TV Gossip Show”

The highest percentage of purchase in 2011 results node 10. It is a person who watches “TV Gossip Show” 2 times a week or less and married 50’s. Their index value is 124.0%. Node 3 is the next highest percentage that is 114.1%. It is a person who watches “TV Gossip Show” 2 or more times, 20’s, 40’s and 50’s. Node 7 is the third highest percentage that is 107.2%. It is a person who don’t almost watch “TV Gossip Show” or watch it 4 or more times a week. Node 8 is the worst percentage that is 58.9%. It is a person who doesn’t almost watch “TV Gossip Show” and 3 or more programs on Tuesday.

In 2013, “TV Gossip Show” surely watches at least once a week and those who have more than 2 or 4 programs on Tuesday tend to have higher purchasing ratios than the whole. Node 8 is the highest value is married with that attribute (112.2%). Node 7 is next highest value is unmarried with that attribute (103.7%). Node 6 is the worst percentage that is 85.2%. It is a person who doesn’t almost watch “TV Gossip Show” and at 20 o’clock.

5.4 When the First Independent Variable Is Set to “Sports”

The highest percentage of purchase in 2011 results node 3. It is a person who watches “Sports” once or more times a week, 20’s, 40’s and 50’s. Their index value is 115.7%. Node 7 is the next highest percentage that is 104.0%. It is a person who watches “Sports” once or more times a week and married 30’s. Node 8 is the worst percentage that is 73.1%. It is a person who watches “Sports” once or more times a week and unmarried 30’s.

The highest percentage of purchase in 2013 results node 6. It is a person who watches “Sports” 3 or more times a week, 20 o’clock program once time or, 5 or more times a week. Their index value is 111.2%. Node 9 is the next highest percentage that is 108.1%. It is a person who watches “Sports” once time a week, “20 o’clock program once time or, 4 or more times a week. Node 5 is the worst percentage that is 86.6%. It is a person who doesn’t almost watch “Sports” and at 20 o’clock.

5.5 When the First Independent Variable Is Set to “Variety Show”

The highest percentage of purchase in 2011 results node 9. It is a person who watches “Variety Show” once to twice times a week, married and one to two, or 4 to 6 programs on Friday. Their index value is 119.7%. Node 3 is the next highest percentage that is 114.2%. It is a person who doesn’t almost watch “Variety Show” or watches 5 or more times and watches “NEWS” twice or more times a week. Node 7 is the third highest percentage that is 101.3%. It is a person who doesn’t almost watch “Variety Show” or watches 5 or more times and watches “NEWS” one time or less a week.

Node 6 is the worst percentage that is 76.4%. It is a person who watches “Variety Show” one e or twice times a week and unmarried.

The highest percentage of purchase in 2013 results node 8. It is a person who watches “Variety Show” 4 or more times a week, married and 18 o’clock program 1 to 6 times a week. Their index value is 111.0%. Node 6 is the next highest percentage that is 103.8%. It is a person who watches “Variety Show”4 or more times a week, unmarried and “Documentary” 3 to 6 times a week. Node 2 is the worst percentage that is 84.4%. It is a person who don’t almost watch “Variety Show”.

5.6 When the First Independent Variable Is Set to “Hobbies and Education”

The highest percentage of purchase in 2011 results node 6. It is a person who watches “Hobbies and Education” once time, or 7 times or more a week, married and 4 programs or more on Monday. Their index value is 143.4%. Node 2 is the next highest percentage that is 118.1%. It is a person who watches “Hobbies and Education” twice to 6 times a week. Node 5 is the third highest percentage that is 101.6%. It is a person who watches “Hobbies and Education” once time, or 7 times or more a week, married and 4 programs or less on Monday. Node 8 is the worst percentage that is 66.9%. It is a person who watches “Variety Show” once a week and unmarried 30’s.

The highest percentage of purchase in 2013 results node 11. It is a person who watches “Hobbies and Education” one or more times a week, married and 22 o’clock program 3 or, 7 times or more a week. Their index value is 115.3%. Node 14 is the next highest percentage that is 112.4%. It is a person who doesn’t almost watch “Hobbies and Education”, 20 o’clock program 3 or, 7 times or more a week and “Sports” once a week. For the third and later, node 5 is 109.6%, node 12 is 103.8%, node 10 is 103.5%. Node 9 is the worst percentage that is 81.3%. It is a person who watches “Hobbies and Education” once a week, unmarried and 23 o’clock program 2 times or less a week.

6 Consideration of the Results and Suggestions

Even though purchasing has increased from about 57% to 77.9%, the maximum value of the index value was about 13.9% lower between 2011 and 2013. In 2011, the result that the maximum value was about 30% different depending on the program category. But in 2013 there was little difference (Table 7). About analysis model, we need to consider the other analysis methods such as linear models. Because correct answer rate is not good in 2011.

Table 7. Maximum index value
Full size table

From the above result, we compare and consider each analysis. When we set the first independent variable to NEWS, in the results of 2011, people watching many categories of programs, such as TV drama, and NEWS’s viewing frequency were slightly lower and married people in their 50’s. In 2013 people became changed that they watch the Variety Show and the time zone is 17 o’clock or 18 o’clock.

When we set the first independent variable to “Documentary”, in the results of 2011, it was many purchased married 50s people. In 2013, in addition to the information on married people, Variety Show and the number of viewing times at 20 o’clock appeared at the node. When we set the first independent variable to “TV Gossip Show”, in the results of 2011, it was many purchased married 50s people. In 2013, in addition to the information on married people, the number of viewing times at Tuesdays appeared at the node. When we set the first independent variable to “Variety Show”, The number of views “Variety Show” has increased in two years. In the result of 2011, the next node that it was Married was the number of viewing programs on Friday, but in 2013 it changed to 18 o’clock. When we set the first independent variable to “Hobbies/Education”, In the result of 2011, it was a married person who watched on Monday but in 2013 it changed to a person who is married and watches at 22 o’clock.

We made a consideration by category. It turned out that factors to be considered in everything changed. We made 12 decision tree analysis in 2011 and 2013. We also mentioned the worst in each analysis results, but about half of them resulted in buyers of chocolate products not frequently watching TV. From this, it was also suggested that TV has an influence on purchasing chocolate products. It was appeared nodes that “married or unmarried” to the all analyzed results. this is an important element of customer information in the future. Nodes separated by age appeared only in 2011. However, the node that splits in broadcasting time zone appeared only in 2013. From this, it can consider that matters to be emphasized have changed. we think it is effective that the broadcasting time zone is better than Age of watching programs. From the analysis of all decision trees, the values of “Hobbies/Education” were better than those of other results in both years. Specifically, the index value in 2011 was the highest. And in 2013, nodes that the index values exceeding 100% was the most. “Hobbies/Education” is a suitable program category for advertising chocolate products. Especially we consider 20 o’clock or 22 o’clock is good.