Using the business intelligence programs that you are already familiar with to extract useful information from customer interactions can turn out to be difficult. Most business intelligence analytics like OLAP (On Line Analytical Processing) and data warehousing function very well when dealing with structured data, but when faced with unstructured data like customer comments these tools are not helpful in and of themselves. The question is how to make customer comments, whether collected through surveys, customer service calls, comment cards, or other means, assessable to these tools. These comments can include information on everything from performance to service quality to cost to reliability. The key is to organize the unstructured data.
In order to structure this data, a few questions have to be addressed:
- What kind of information is your company looking for in the comments?
- What categories should the comments be broken down into?
- What kinds of terms are important to categorize these comments?
The types of categories for a comment can usually be broken down into complaints, either about the product, delivery, shipment, and/or cost; questions about the product, delivery, or billing; compliments about the product or service; and suggestions for how to improve service. Using techniques for text mining, you can find the most important terms and phrases in the comments to add a new level of analysis to your current business intelligence findings. This process will find the names of products or people, dates, times, and monetary amounts. There is a variety of entity extraction tools on the market, including a number of open source versions that can be starting points for your company to build a custom version.
Nouns, verbs, and adjectives will all be identified and can be used to identify the type of comment and its tone. The comment can then be categorized and the unstructured data begins to take on a structure that can be analyzed by your business intelligence program. Using fixed categorization, the comments can be used in existing structured data elements in mining and statistic analysis.
Since the words and terms will vary from customer to customer and region to region, the different versions of the same concept will have to be recognized as one and the same and processed as such by the tools that you use. The same applies to semantic rules. Once you find a text mining technique that can deliver the structured information that your company can use to make strong decisions based on business intelligence you can rest assured that the information you have from your customers could be used to enhance your business.