Contact Details

 

www.StatAdvice.com
info@StatAdvice.com

Skype login: StatAdvice

  • We are available 7 days a week

  • We offer free initial consultation

  • You don't have to pay before we send you a preliminary report

 



Consumer reviews mining: doing actionable quantitative
market research without survey data

With the growth of e-commerce and consumer reviews websites quantitative market research and text mining techniques can now be applied to non-survey data. Using such data is cheaper and by far more actionable than analyzing stated preferences data or even conjoint analysis data obtained in laboratory settings. It is also very fast. For instance, it takes only a week to collect 300 reviews, process the data, and prepare a detailed report.

 

Reinhold Decker (Bielefeld University) and Michael Trusov (University of Maryland) were among the first who did text mining of mobile phones consumer reviews. In their paper on estimating aggregate consumer preferences from online product reviews published in Intern. J. of Research in Marketing in 2010 they provided a framework for measuring how mentioning various positive and negative features in text reviews influences the satisfaction rating, given by consumers. Approximately at the same time we started applying similar techniques to data, which was interesting for internet stores and manufacturers.

 

We see several purposes for which consumer reviews mining is extremely useful:

  • A new model or a modification of an older model is being developed and you want to figure out the advantages and disadvantages of the previous models, as well as the functional attributes that are the key drivers of satisfaction and dissatisfaction

  • For new product development you need to understand what is valued by the market. You need general information about the key drivers of satisfaction on the market for your product (mobile phones, laptops, smartphones, etc.) Then you should use a pooled sample of reviews for several key models. This will allow both key drivers analysis plus comparison of competing products. In the case of multiple product analysis it is possible to conduct correspondence analysis (which results in perceptual maps) to identify characteristic features of each product and/or brand.

  • You want to do comparative analysis and see how your products are perceived by consumers compared to the competition.

The key problems with consumer reviews mining are:

  • Natural language processing when slang and mistakes are common in the text of reviews

  • Feature importance measurement

  • Fraud detection

Our experiments have shown that the highest quality is achieved when experienced people (and not just text mining software) look for the sentiments in reviews. This is especially true in the case of the Russian language, for instance. In addition machines are not good at dealing with spelling errors and informal language, which are common in the text of reviews.

 

Although the analysis is based on data mining and statistical analysis, we recommend avoiding the full automation of knowledge discovery. Instead, we scrutinize data as much as possible to get the most out of data. In addition, we filter out suspicious reviews (such as those with the order ID in the text of the review). Several techniques are used for feature importance measurement, including regression and classification trees, Shapley Value Decomposition and other regression-based methods.

 

This cutting-edge technique is not widely offered by research companies, but at StatAdvice.com we’ve already completed several such projects with consumer electronics manufacturers and internet stores. 

 

Since our mission is to disseminate cutting-edge marketing science techniques among marketers, we set affordable prices for consumer reviews mining projects. For more information on what our consumer reviews mining offer includes, please follow the link:

http://www.statadvice.com/proposalconsumerreviewseng.pdf 

 

Eugene Antipov

Market researcher and Statistical Consultant

e-mail: info@StatAdvice.com

www.StatAdvice.com

 

Our clients usually find us when they search for:

dissertation statistics consultant, dissertation statistics consulting, dissertation statistics, dissertation help service, dissertation help uk, online dissertation help, dissertation help service, help in statistics, help on statistics, college statistics help, online statistics help, help with statistics, spss help, statistical analysis help, phd help desk, phd thesis help, phd dissertation help, dissertation consulting, statistical analysis consulting, statistical consulting services, statistical consulting center, statisticsl services, statistics help, help with spss, help with stata, help with eviews, help with statistica, statistical consultant, data mining, Statistical analysis, statistical consulting, statistical analysis and statistical consulting, data mining, SPSS, EViews, Statistica, Excel, Stata, market research, marketing research, survey data analysis, Monte-Carlo simulation, Movie sales projections, star power, star power ratings