As an emerging voice of the customer (VOC) containing feedback such as opinions and expectations about products, social media data have the potential used for product improvement and new product development. However, most prior studies have focused on determining customer concerns, while neglecting to incorporate them into a systematic approach to identify product opportunities. In response, this paper suggests an approach to identify product opportunities from customer reviews in social media. This approach employs topic modeling to identify the product topics discussed by customers from large-scale review posts related to a given product. A keygraph is then constructed based on the co-occurrences among the topics contained in each post. The chance discovery theory is then applied to generate new product opportunities from the chance nodes obtained from the keygraph. Our approach contributes to the systematic ideation process for product opportunity analysis based on large-scale and real-time VOC.
The first step in our approach involves collecting online customer reviews of a given product as the material for analysis. In collecting such review data, analysts should consider two factors: the type of product and the type of online data. First, the products applicable to this approach are the high-tech products that contain various functions, components, and accessories. The majority of recent online product reviews in social media are usually generated for high-tech products because these products, which rapidly evolve and have short lifecycles, can receive a variety of feedback related to their functions, components, and accessories; as the number of product reviews that are used increases, our textual analysis becomes more effective.
In this step, LDA-based topic modeling is employed to identify product topics that are being discussed by product customers in social media (Figure 4). Our LDA-based topic modeling requires two inputs: a document-keyword matrix and the number of topics. The document-keyword matrix in the previous step can be used as the input matrix for topic modeling, while an appropriate number of topics should be determined for topic modeling. Several techniques can be used to select an appropriate number of topics, including the elbow method, information criterion method, and information theoretic method. Of these methods, the elbow method is used in this study, in which an optimal number of topics is determined by calculating the average cosine similarity value between pairs of topic-word distribution vectors outputted by topic modeling
In this step, a keygraph is generated that visualizes an overall landscape of the relationship among product topics and shows the breaking points of the graph (Figure 5). As previously mentioned, the keygraph generation process requires two parameters for its inputs: the co-occurrence data of item-pairs and the number of chance items. In this step, the co-occurrence between pairs of product topics can be identified by analyzing the document-topic distribution matrix produced in the previous step. Each row vector of the document-topic distribution matrix shows the degree to which a document belongs to all product topics in terms of probability.
In the recent competitive business environment, various firms have been attempting to continue improving their product business to cope with rapidly changing market trends and customer needs. In this regard, the ability to identify product opportunities, which are defined as a chance to develop new products or improve current products, is considered to be most essential for the sustainable growth of product-based firms .
To effectively identify such opportunities, in product planning processes, the voice of the customer (VOC) (which includes customer expectations and opinions about a product) is generally considered the primary prerequisite . Collection and in-depth analysis of the VOC of products can help firms determine product development directions in a more practical and reasonable way , thereby eventually enabling firms to build a customer relationship that cannot be easily copied by their competing firms.
this approach to product opportunity identification using social media mining involves the following steps: 1) gathering online customer reviews related to a specific product, 2) extracting the product�s discussion topics from the online customer reviews by applying topic modeling, 3) constructing a keygraph based on the co-occurrences between pairs of the discussion topics in the online reviews, and 4) generating product opportunities based on chance discovery analysis that uses breaking topics and their neighboring topics in the keygraph.
To illustrate the operation of our product opportunity analysis approach, we apply it to the online posts of the Samsung Galaxy Note 5. The contribution of this study is two-fold. First, this approach will contribute to the systematic identification of new product opportunities from large-scale and real-time customer feedback in social media while being a useful aid for monitoring rapidly changing customer needs and relationships among these needs. Second, this approach is a tool for product opportunity ideation that will have a synergetic effect when incorporated into the product planning process.
VOCs are known to be an excellent material for the process of product development. Recently, as customer voices have accumulated and been shared through various social media, many firms have attempted to utilize social media as a tool to improve their competitiveness. To this end, various social media data analyses have been conducted. However, these studies mostly centered on identifying current product trends and their relevant sentiment or opinion. In addition, while they focused only on the customers� main interests, they did not lead to exploring potential product opportunities.
Therefore, an approach was proposed in this paper to identify the opportunities of a specific product through social media mining by combining topic modeling and chance discovery theory. In the steps of this approach, we used topic modeling to find what product topics customers are interested in and chance discovery theory to create new product opportunities that may exist around breaking product topics.