Using Qubole Data Service on Azure to analyze retail customer feedback

Posted on 4 January, 2018

C+E Azure Architect, Global black belt

It has been a busy season for many retailers. During this time, retailers are using Azure to analyze various types of data to help accelerate purchasing decisions. The Azure cloud not only gives retailers the compute capacity to handle peak times, but also the data analytic tools to better understand their customers.

Many retailers have a treasure trove of information in the thousands, or millions, of product reviews provided by their customers. Often, it takes time for particular reviews to show their value because customers "vote" for helpful or not helpful reviews over time. Using machine learning, retailers can automate identifying useful reviews in near real-time and leverage that insight quickly to build additional business value.

But how might a retailer without deep big data and machine learning expertise even begin to conduct this type of advanced analytics on such a large quantity of unstructured data? We will be holding a workshop in January to show you how easy that can be through the use of Azure and Qubole’s big data service.

Using these technologies, anyone can quickly spin up a data platform and train a machine learning model utilizing Natural Language Processing (NLP) to identify the most useful reviews. Moving forward, a retailer can then identify the value of reviews as they are generated by the user base and gain insights that can impact many aspects of their business.

Join Microsoft, Qubole, and Precocity for a half-day, hands on lab experience where we will show how to:

  • Leverage Azure cloud-based services and Qubole Data Service to increase the velocity of managing advanced analytics for retail
  • Ingesting a large retail review data set from Azure and leverage Qubole notebooks to explore data in a retail context
  • Demonstrate the autoscaling capability of a Qubole Spark cluster during a Natural Language Processing (NLP) pipeline
  • Train a machine learning model at scale using Open Source technologies like Apache Spark and score new customer reviews in real-time
  • Demonstrate use of Azure’s Event Hub and CosmosDB coupled with Spark Streaming to predict helpfulness of customer reviews in real-time

This workshop can be the basis of creating business value from reviews for other purposes including:

  • Fake review fraud detection
  • Identifying positive product characteristics
  • Identify influencers
  • Uncover new feature attributes for a product to inform merchandising

Register today for our event in Dallas, Texas on January 30th, 2017.

Space is limited, so register early!