BMW 7 Series in the Chinese social web - Anacodeanacode.de/wordpress/wp-content/uploads/2018/12/BMW7-Social-Re… · Figure 1: Frequent concepts for BMW, Audi and Mercedes-Benz in

BMW 7 Series in the Chinese social webA comparison with Audi A8 andMercedes-Benz S-Class

Content

Concepts and topics

Evolutions in time

Customer sentiment

Key influencers

Dr. Janna Lipenkova, Tomas Stibrany, De NingJanuary 2017

BMW 7 Series in the Chinese social web:A comparison with Audi A8 and Mercedes-Benz S-Class

Dr. Janna Lipenkova, Tomas Stibrany, De Ning

January 2017

Contents

Executive summary 3

1 Introduction 4

2 Concepts and topics 7

3 Evolutions in time 10

4 Customer sentiment 11

5 Key influencers 15

6 Outlook 16

2

Executive summary

Figure 1: Frequent concepts for BMW, Audi and Mercedes-Benz in Chinese social media

This report provides a descriptive overview of the Chinese Web 2.0 landscape for automotive feedback,focussing on BMW 7 Series and comparing it with Audi A8 and Mercedes-Benz S-Class. The feedbackis analysed both from a qualitative and a quantitative perspective. The main observations and findingsare as follows:

Channels Different data channels and web resources have different degrees of specialisation and so-cial reach – in general, the level of specialization is antiproportional to social reach. In order to geta comprehensive overview over social feedback, multiple resources should be combined and leveragedalong these criteria. The present report considers data from forums, product reviews and social networks.

Popular topics and concepts We find that users are most concerned about the price and optical aspects(design, visual appearance) of the three considered series. Competitor brands that are discussed in acomparative perspective are mostly high-end or consumer-oriented foreign brands from Germany, USand Japan, whereas native Chinese brands are much less frequent. Geographically, users concentrate inthe big cities and more affluent regions along the East coast.

Temporal evolutions The quantity of buzz grows relatively evenly for all three series before 2015,with BMW 7 and S-Class leading. In 2015 – 2016, there is a burst in the quantity of data for BMW 7,which correlates with the introduction of the new generation of the series.

User satisfaction and sentiment Users are generally satisfied with the frequently mentioned majorproduct features of BMW 7. There are, however, some categories that are perceived negatively – specifi-cally, components related to the front part of the car, the fuel consumption and aspects related to acousticquality and insulation.

Social influencers Among the key influencers on WeChat, China’s leading social network, we mostlyfind media accounts posting on general automotive topics. There are no accounts with a wide socialreach that would specialise on the BMW brand. Thus, influencer marketing is an opportunity yet to beexplored by BMW’s marketing and branding strategy.

BMW 7 Series in the Chinese social web 3

1 Introduction

1.1 Goals

With the quick rise of digital service channels in the Chinese automotive industry, online customerfeedback has increased both in quantity and in authority and forms a large data basis for extractingcustomer insights. The present report provides a synopsis of the feedback on BMW 7 Series that isshared by Chinese customers online and compares it with data for Mercedes-Benz S-Class and AudiA8. The research is based on three different types of social data – forum discussions, product reviewsand social network posts. The following questions are addressed in a comparative perspective:

• What: what are the main topics that users talk about with respect to the three series?

• When: how does interest in the three series evolve over time?

• How: which sentiments and opinions do users express?

• Where: what are the channels via which the products are discussed? What are the distinguishingcharacteristics of these channels?

• Who: who are the ‘opinion leaders’ in the field?

1.2 Tools

Two analysis tools were used for producing the analyses in the report:

• Anacode Web&Text API (https://api.anacode.de)

• Anacode aggregation and plotting toolkit (https://github.com/anacode/anacode-toolkit)

The data set and code for the analysis of product reviews is made available at this link: https://api.anacode.de/api-docs/bmw7 review comparison.html.

1.3 Data

To provide a comprehensive overview, the data sample used for this report combines three different datatypes:

1. Forum discussions:

• http://autohome.com.cn

• http://pcauto.com.cn

• http://auto.163.com.cn

2. Product reviews:

• http://car.bitauto.com

• http://autohome.com.cn

• http://www.58che.com

• http://pcauto.com.cn

• http://koubei.16888.com

• http://www.niuche.com

3. Social network data:

BMW 7 Series in the Chinese social web 4

• https://www.wechat.com

Each data type has specific characteristics and is thus suitable for extracting different kinds of in-sights. In particular, four criteria are relevant when considering data for social text analytics:

1. Level of detail: how specialized is the data? Do users talk about brands and series in general or dothey also go into more fine-grained attributes of the products, such as design, specific components,performance etc.?

2. Level of noise: how much of the data is ‘clean’ text and can be analysed in a reliable way usingtext mining? How much is advertising or spam?

3. Social reach: how high is the visibility of the data? Who consumes the data - is it a delimitedgroup of enthusiasts or is the general public also exposed to it?

4. Brand/product focus: is each datapoint focussed on a specific brand or model, or does it cover alarger set of topics?

Table 1 depicts these and some additional properties of the considered data types. From a quantitativeperpective, figures 2 and 3 show the data quantities for forums posts and product reviews. For socialnetwork data, we use WeChat as only resource. The data sample consists of around 20 millions of postsby public accounts (公众号) produced during December 2016. Of these, 84,979 posts contain mentionsof BMW as a brand, and 3,191 posts mention BMW 7 Series.

Table 1: Characteristics of the considered data channels

BMW 7 Series in the Chinese social web 5

Figure 2: Quantity of forum posts per domain and series

Figure 3: Quantity of product reviews per domain and series

The following sections present analyses which leverage the specific characteristics of each data type:

1. Forum data are used to produce a conceptual overview of what users talk about and embed it intoa geographical and temporal context.

2. Product reviews are used for qualitative sentiment analysis and evaluation on feature level, thusshowing how users think and feel about fine-grained aspects of the products.

3. Social network data are used to identify opinion leaders, i. e. influential users with expert knowl-edge in the topic.

BMW 7 Series in the Chinese social web 6

2 Concepts and topics

The concept cloud in Figure 4 presents the data from a conceptual point of view. The size of eachconcept in the cloud corresponds to its frequency and thus relevance in the data. As expected, the conceptAutomobile clearly stands out. It is surrounded by the three considered brands (BMW, Mercedes-Benz,Audi), major product attributes (Visual Appearance, Comfort, Configuration etc.) and price-relatedconcepts (Price, Bargain Price, Low Price).

Figure 4: Conceptual overview over forum data

To provide a more detailed picture, we consider frequencies for three specific types of concepts:

• Brands (Figure 5): beyond the brand whose product is actually evaluated, reviews often alsomention other comparable brands. This information can be used for competitive intelligence. Asexpected, for each of our series, its own brand is the most relevant. The list of remaining brandsthat are mentioned includes high-end brands, such as Maybach and Rolls Royce, but also moreconsumer-oriented lines such as Volkswagen and Toyota. It is worth noting that the list does notcontain any domestic Chinese brands.

BMW 7 Series in the Chinese social web 7

Figure 5: Frequent brands in forum data

• Features (Figure 6): features and attributes break a product down into those attributes that arerelevant to the users. Features can be both ‘abstract’ attributes, e.g. Visual Appearance andSafety, and concrete components, such as Engine and Seats. We see that Price is at the top forall three series, thus reflecting the importance of the cost factor even in the premium segment.It is followed by Visual Appearance for BMW 7 and S-Class; in the case of Audi A8, VisualAppearance competes with Comfort. The importance of visual matters confirms the long-standingrole of the car in China as a status symbol, not to say an instrument for ‘showing off’.1 Beyondthis, the list is mostly composed of larger user-facing components (Interior, Body, Seats etc.) aswell as performance-related characteristics (Car Engine, Tires).

Figure 6: Frequent features in forum data

• Locations (Figure 7): geographical locations are relevant from two perspectives: first, they oftenreveal the location at which the user uses or purchases the product. Second, they show howsensitive users are to the origin of a product. For instance, it is well-known that the ‘Made inGermany’ label is an important selling point for cars and other tech products in China due to theinherent respect of Chinese people for German engineering quality. On the international level,China as the country of most users and Germany as the country of origin of the products lead thelist for all three series, followed by USA and Japan. Looking into the breakdown for China, wesee that most of the provinces and cities are located in the more affluent coastal regions. The mapin Figure 8 further depicts the distribution of mentions on the national level.

1However, as found by McKinsey, the role of cars as status symbols is currently decreasing in China, with budget andpractical considerations taking the lead (source: McKinsey China Auto Consumer Report 2016).

BMW 7 Series in the Chinese social web 8

Figure 7: Frequent locations in forum data

Figure 8: Distribution of mentions among Chinese provinces

BMW 7 Series in the Chinese social web 9

3 Evolutions in time

The forum format encourages spontaneous discussion by users; thus, the quantity of forum buzz can beused as an indicator for the general interest in the topic. Figure 9 shows how post quantities for the threeseries evolved over the past twelve years. We see that, while the developments are fairly similar for thethree series before 2015, BMW clearly outgrows the other two series after 2015. This can be explainedby the introduction of the 6th generation of BMW 7 Series in 2015 – 2016, which also comes withproduct configurations adapted specifically to the Chinese market. Figure 10 gives a more fine-grainedview by depicting the monthly evolution starting from Q4 2015.

Figure 9: Forum data quantities in time (2004 – 2016)

Figure 10: Forum data quantities in time (Sept 2015 – Nov 2016)

BMW 7 Series in the Chinese social web 10

4 Customer sentiment

In Section 3, we have studied what users talk about when discussing the three series. In this section, wewill study how they talk about it. Specifically, we will go into more detail about features – those productaspects that are frequently discussed and evaluated by the users and thus are relevant to them. Featurescan be both physical components, such as the Engine and the Seats, and ‘abstract’ features, such asthe Price and the Comfort. The concept cloud in Figure 11 illustrates those features that are frequentlydiscussed for BMW 7 Series.

Figure 11: Features that are frequently discussed in BMW 7 product reviews

In the following, we consider user sentiments for specific features from two points of view:

Frequently evaluated features Figure 12 shows the sentiment values for most frequently evaluatedcomponents. We see that the major aspects that determine the inner configuration of the car – Body,Space, Seats etc. – are evaluated positively, also in comparison to the other two series. Only two of thecomponents – Head Light and Interior – get rather negative evaluations and also are outperformed byboth Audi A8 and S-Class.

Figure 13 shows sentiments for ‘abstract’, non-component features that are most frequently evalu-ated for BMW 7. Again, we see that most of the popular attributes receive rather positive evaluations,with the exception of Fuel Consumption and Price. Comfort, although rated quite well, still scores lowerthan for the other series. The evaluations of Price are generally not very high, with BMW 7 receivingthe lowest value. The following examples from the original data illustrate these problematic features:

(1) a. 油耗有点点高

‘The fuel consumption is relatively high.’

b. 前排座椅不够舒适

‘The front seats are not comfortable enough.’

c. 便宜点就好

‘A cheaper price would be welcome.’

BMW 7 Series in the Chinese social web 11

Figure 12: Sentiments for components with most evaluations for BMW 7

Figure 13: Sentiments for abstract features with most evaluations for BMW 7

BMW 7 Series in the Chinese social web 12

Contrasts between the series Figures 14 and 15 show those features for which BMW 7 shows strongpositive and negative divergences with the other series. Beyond an outstanding price-quality ratio forBMW 7, features related to the rear components of the car (Rear End, Rear Light) are also favored bythe users:

(2) a. LED尾灯看起来时尚

‘The LED rear light looks very stylish.’

b. 尾灯的效果也非常棒

‘The quality of the rear light is also very good.’

c. 尾部很大气

‘The rear side is very impressive.’

By contrast, front components (Front Seat, Head Light) receive more negative evaluations. Finally,acoustic features (Volume, Acoustic Insulation, Multimedia Sound) are also perceived negatively forBMW 7:

(3) a. 大灯外观太模糊

‘The head light looks too vague.’

b. 感觉头部不适

‘My impression is that the front doesn’t fit well.’

c. 音响音质一般

‘The sound quality is quite ordinary.’

Figure 14: Features with distinctive positive evaluations for BMW 7

BMW 7 Series in the Chinese social web 13

Figure 15: Features with distinctive negative evaluations for BMW 7

BMW 7 Series in the Chinese social web 14

5 Key influencers

As mentioned in Section 2, WeChat data have extremely high visibility and social reach among Chineseweb users.2. Thus, this data is suitable for identifying people and social accounts that have a high levelof expertise and influence on a given topic and thus can be used for the dissemination of informationabout a brand and its products.

We use our WeChat data to filter out those authors that produce a certain amount of content onBMW. Figure 16 shows twenty of the top authors, along with the quantity of produced content (bothcontent about BMW and other content) and their influence metrics (average reads and likes per article).All accounts have more than 4,000 average reads. Most of these accounts are media companies or self-media specialised on the automotive domain; it is notable that none of them is focussed uniquely onthe BMW brand: while WeChat has BMW-focussed accounts (e. g. BMWsky宝马会,周口周德宝马),these accounts have a relatively low social reach and thus cannot be taken into account as significantinfluencers.

Figure 16: Key influencers for BMW on WeChat

2In Q3 2016, WeChat counted 846 millions active users (https://www.statista.com/statistics/255778/number-of-active-wechat-messenger-accounts, last visit: Jan 5, 2016.

BMW 7 Series in the Chinese social web 15

6 Outlook

While our initial findings mostly confirm and complete known facts about the Chinese market and pro-vide series-specific insights, they do not fully uncover the combined potential of web data and textanalytics technology. A number of extensions and deep dives will help to reach a level of detail fromwhich concrete actions can be derived. With respect to the data, the following enhancements can beapplied:

• Differentiation of perceptions by stakeholder groups: beyond adding more web resources ofthe considered data types (product reviews, forums, social networks) to improve the statisticalcoverage of the results, additional web data types can provide a new perspective on the market.Specifically, professional web data such as articles in newspapers and web portals can uncoverhow the brands and products are perceived and presented on a more public level. Resources thatare managed by the brands themselves (company websites, social accounts etc.) can be used toanalyse the official brand and product messaging.

• Customer profiling: the present research mainly focussed on the extraction of relevant conceptsand opinions unstructured content. However, the studied data also contains structured informationabout the users. For instance, most product review sites ask users to provide the location wherethe car was purchased, as well as the intended uses of the vehicle. These bits of information canbe combined into profiles of specific customer segments, thus providing an additional dimensionfor the content analysis.

The text and content analysis can be extended as follows:

• Decomposition into fine-grained concepts: in the report, concepts and topics were presented asatomic entities. In the real world, each of them can be further split into more granular concepts –for instance, whereas the feature Visual Appearance is a very general starting point, users normallytalk about the visual appearance of specific components, such as the lights, the interior design,the dashboard etc. Our text analytics technology allows us to analyse these associations andalso provide the sentiment for each association. From a practical perspective, this fine-graineddecomposition is particularly useful for topics and features that are most discussed and / or mostproblematic.

• Temporal and geographic analysis: in Sections 3 and 4, we have provided high-level analysesof the geographical and the temporal contexts of the data. Just as for fine-grained content analysis,it is possible to “zoom in” on specific topics and analyse their geographical and temporal distri-bution. Temporal breakdowns allow to spot upcoming trends and to see how marketing events(launches, campaigns etc.) influence the interests and the focus of the users. The geographicalperspective will provide hints for the localisation of product, marketing and sales strategies.

• Explorative analysis and trend discovery: in the present analysis, we have used our existingconsumer insight ontology. However, new trends and product features are constantly under way– especially in capricious, quickly evolving markets such as China. These emerging concepts canbe analysed using explorative algorithms, which, instead of using a predefined set of concepts, aredriven by the data and allow to uncover completely new topics.

BMW 7 Series in the Chinese social web 16