Business Intelligence to Predict the Usefulness of Online Reviews

1. Theme:Identifying the ‘helpfulness’ of online customer textual reviews to make business decisions.

User-Generated Content (UGC) help to reduce the perceived risk of online shopping when customers decide to make any purchase decisions. However, with the increase of online reviews posted in social media and online shopping websites is virtually impossible for costumer to read all the reviews before making any purchase decisions, especially for products that have been reviewed hundreds of times. Classify such a large volume of unstructured data that can be presented easily for the customer’s eyes has become a management challenge.

Using Business Intelligence to predict and analyse the ‘helpfulness’ of online customer textual reviews would be an excellent implementation to facilitate customer’s final purchasing decisions and reduce marketing costs by better understanding customer preferences and implementing market segmentation. It is useful to understand the perceived helpfulness of online reviews, as helpful online reviews play an important role in purchasing decisions.

2. Scope and boundaries:

Scope: Apply business Intelligence for online e-commerce in customer textual reviews that allows making business decisions such as, segmentation of customers and selection of the best products for them (marketing), managing inventory according to costumers demand (logistic) and web merchandiser (merchandiser).

Boundaries: The study will be implemented in firms where their primary business is sale and purchase of goods on an online platform, such as Amazon, eBay, Iconic. Other kinds of business such as manufacturing, and services are excluded from the scope of this study.

3. Business functions:

Data provides insights about customer behaviours and businesses make use of those insights for market intelligence and to bring strategies for their business functions. In an e-commerce platform main business functions are represented as follow:

A) Marketing: Analysing reviews and categorizing them to give support in online purchasing would facilitate the marketing team. The selection of products that have been popular in the business and reducing inventory in those products that don’t exceed customer’s expectations.

Online review helpfulness could be influenced positively by the selection of the best products and to generate marketing strategies that support the growth of the business.

B) Logistic: Knowing the best products, segmentation of customers and number of sales are the key elements to provide an excellent logistic process such as inventory.

C) Web merchandiser and web design: How the products are organized and how an e-commerce business designs their websites are key elements to sell their products and target customer according to their preferences and analysing their reviews. With a deep understanding of customer’s comments/ feedback, the business strategy would have a better implementation and a business’ profit increasing.

4. Scenarios

Understanding customer reviews will help to create business functions and create business growth. Data from customer reviews are a powerful tool to align business strategies, Identifying the ‘helpfulness’ of online customer textual reviews bring the facility for customers to analyse and reduce online purchasing risk and generate data that can be used to take business strategies.

Possible scenarios would be where marketing, logistics and web merchandiser and web design interact would be:

Image 1. Conceptual model and possible scenarios

Marketing:

-          Segmentation costumers and select the best products of them

-          Adding new lines of products that satisfy customers’ needs

-          Managing promotional campaigns, such as sale products and promotional channels.

-          Setting prices and create new promotion strategies.

Logistic:

-          Managing inventory according to customer demands. Creating inventory strategies

-          Managing procurements of goods and services

-          Managing key partners: Managing warehouses, packing, storage and transport

Merchandiser:

-          Web merchandiser of products according to customer preferences

-          Create high visibility and classify products according to customers’ demands

5. Existing use of BI:

BI has become increasingly important for firms’ profit and operations. The numerous customer online reviews posted on social media and online shopping websites have sped up the demand for big data analytics and corresponding techniques. Dealing with unstructured texts is among one of the biggest challenges of big data analytics and different firms are adopting techniques to analyse and take decisions from customer’s reviews. Some examples are:

1. ‘Business intelligence in online customer textual reviews: Understanding consumer perceptions and influential factors’ (article).

This study examined costumer satisfaction and dissatisfaction towards attributes of hotel products based on online customer textual reviews. With techniques such as text mining approach, latent semantic analysis (LSA), they identify the key attributes driving customer satisfaction and dissatisfaction toward hotel products and service attributes. This study bridges customer online textual reviews with customers’ perceptions to help business managers better understand customers’ needs through UGC (Xun Xua et al. 2017).

2. ‘Mining customer requirements from online reviews: A product improvement perspective’ (article).

Extracting data from online reviews facilitated the study for a manufacturing firm to create product development. In the research, they proposed an automatic filtering model to predict the helpfulness on online reviews from the perspective of the product design. The method such as KANO was implemented in the study. It implementation brought robust data to develop appropriate product improvement strategies. Using information from JD.com (one of the largest electronic market places in China) and applying big data classic management models they conclude that they are excellent models to bring success for a big data commerce (Jiayin Qi et al. 2016).

6. BI practices

Using different practices implemented in the studies shown above, the methodology that I will implement in my BI analysis is as follow:

1. Data collection and pre-processing: Review data collection and select the appropriate data for the study such as, study market and reviews necessary to analyse. Analysing unstructured data to translate and extract customer’s valuable information.

Using LSA, understanding for LSA as an algebraic-statistical method that can detect the underlying topical structure of a document corpus and extract the hidden semantic structures of words and sentences (Evangelopoulos, 2011). LSA will be applied in three steps: pre-processing, term frequency matrix transformation, and singular value decomposition (SVD).

2. Identifying helpfulness in customer reviews: Usually, reviews are posted independently by customers immediately after purchase, providing firms with accurate feedback, however, there is a large volume of online reviews of varying quality. To extract a piece of valuable information from online reviews is necessary to measure the helpfulness and group those reviews that would help in identify business strategies. The helpfulness prediction will be executed as follow:

Step 1: Helpfulness rating. A Randomly selection of 20-30 reviews of the set of data to be analysed. Every review is scored on a 5-scale, from 1 to 5 by a team of experts. Where 1 representing the least helpful and 5 representing the most helpful.

Step 2: Model selection. Features will be designated according to the reviews selected. The features will be categorized and the group of experts will evaluate them according to their helpfulness.

Step 3. Significance analysis and helpfulness prediction. The significance of the features is analysed, and the significant features are chosen for the helpfulness prediction.

3. Factors (attributes) identified: With the results of LSA is easy to understand factors that would help to identify positive evaluations (positive features such as, excellent quality, price promotion and so on) and factors of negative evaluation (negative features such as, bad quality, price promotion and so on).

4. Text regression: Identifying the independent variables and dependent variables to run the text regression and connecting their relation, through regression vector space. Thus, will be easy to understand how changes in one variable would affect another variable for example, an increase in price, would have a big impact on the quality of products.

Image 2. BI Model

7. Explore other possible usages of

1. ‘Predicting the “helpfulness” of online consumer reviews’ (article).

Due to the large volume of data constantly being generated it can be considered a big challenge for both online business and consumers to understand. It can be a big challenge for customers to go through all the reviews to make purchasing decisions. Using machine learning to analyse the helpfulness of customer reviews using several textual features such as polarity, subjectivity, entropy and reading ease can be helpful methodologies to be implemented in my study. The results of ‘Predicting the “helpfulness” of online consumer reviews’ brought an easy way to understand reviews, helped buyers to write better reviews and improvement of business websites (Jyoti Prakash et al. 2017).

2. ‘Discovering business intelligence from online product reviews: A rule-induction framework’ (article).

New automated tools have emerged to analyse online product reviews, However, most lack the capability of extracting relationships between the reviewer’s rich expressions and the customer ratings. This problem is addressed with the development of a new class of BI systems based on rough set theory, inductive rule learning, and information retrieval methods (Wingyan Chung et al. 2012). Using these methods is easy to extract the relationship between customer ratings and their reviews. The result using them produce high accuracy and rules with high support and confident values helping market sentiment analysis and e-commerce reputation management.

Image 3. BI Methodologies

References

1. Xun X. Xuequn W., Yibai L. Mohammad H. 2017, ‘Business intelligence in online customer textual reviews: Understanding consumer perceptions and influential factors’, International Journal of Information Management, vol. 37, pp. 673-683. (Xun Xua et al. 2017)
2. Jiayin Q. Zhenping Z. Seongmin J. Yanquan Z. 2016, ‘Mining customer requirements from online reviews: A product improvement perspective’, Information & Management, vol. 53, pp. 951-963. (Jiayin Qi et al. 2016) .
3. Jyoti P. Seda I. Nripendra P. Yogesh K. Sunil S. Pradeep K. 2017, ‘‘Predicting the “helpfulness” of online consumer reviews’, Journal of Business Research, vol. 70, pp. 346-355. (Jyoti Prakash et al. 2017)
4. Wingyan C. Tzu-Liang Tseng. 2012. ‘ ’Discovering business intelligence from online product reviews: A rule-induction framework’, Expert Systems with Applications, vol. 39, pp 11870-11879. (Wingyan Chung et al. 2012)
5. Evangelopoulos, N. 2011. ‘Tracing Taylorism’s technical and sociotechnical duality through Latent Semantic Analysis’. Journal of Business and Management, vol, 17, pp 57–74.

Australiaeast

https:// australiaeast.api.cognitive.microsoft.com/text/analytics/v2.1/sentiment