Affiliated with:

121 7

The business value of text is immense, but it is not well understood by most organizations. Analysis of textual data can provide a wide variety of benefits.

It is intuitively obvious to many people that there is great business value that can be derived from the analysis of text. But the relationship between business value and the analysis of text is not obvious to everyone. So, to explain the relationship let’s look at an example.

Once upon a time there was a restaurant chain that had lots of locations scattered around the US. On a monthly basis the restaurant chain would get feedback as to their operations. They would get thousands and thousands of customer feedback replies – messages – each month. Each month these pieces of feedback were printed out and placed on the desk of a vice president. Then next month the replies would be printed out and placed on top of the previous month’s replies. In a month’s time the vice president did not even glance at what was being said to the restaurant chain.

When asked about this the vice president replied, “There’s way too many messages for me to sit down and actually read and absorb each one.”

There was an opportunity to understand what the likes and dislikes of the customer were in those messages. There is the “voice of the customer” that is contained in each message. And the restaurant chain did not have time to even glance at them.

The messages covered a wide range of data and information. The messages talked about such things as:

  • Which dishes were too salty. Too hot. Not big enough. Missing.
  • How rude the waiter or cashier was. (or how nice the waiter or cashier was)
  • Was the table clean when the customer sat down
  • Was the service prompt
  • Was the parking lot adequate

And so forth.

In a word the messages received covered every aspect of the restaurant its food and its ambience.

One day the messages were read into textual disambiguation and processed. A data base was created and the results were visualized. One of the visualizations looked like Figure 1:

121 8

Figure 1: Message Analysis – Occurrences by Topic and Sentiment

In order to understand Figure 1 an explanation is required.

There are seven categories of information found in Figure 1 – other ingredients, people, place, price, process, product, and promotions. Other ingredients refers to items on the menu that are not part of the entree, such as potatoes, bread, soups, salads, and so forth. The category – people – refers to waitresses and waiters, cashiers, busboys, and so forth. Place refers to location and ambience. Price refers to the menu price of the item. Process refers to the process of seating the customer, taking the customer’s order, serving the meal and presenting the check. Product refers to the entrée as listed on the menu. And, promotions refers to the promotions conducted by the restaurant chain. All this context is considered as metadata for the actual results.

The portions of the feedback in Figure 1 that are in red are negative feedback. The positive portion of the feedback are in purple. It appears that there is much more negative feedback than positive feedback. However, this ratio is normal. If a person goes into a restaurant and has a negative experience, the person is likely to let management know. But if a person goes into a restaurant and has a positive experience, then that is what the customer expects. Therefore a ratio of 85% negative and 15% positive is what is expected under normal circumstances. Only if the ratios start to go beyond these ranges is there a perceived problem at the restaurant.

So the positive/negative ratio is about as expected.

However there is some really interesting information in the analysis. Hardly anyone said anything about pricing. The message is that management is probably “leaving money on the table”. Management could be charging a nickel or a dime more for each dish and the customer wouldn’t care. The message that management is not charging enough is a very important message. This is a perfect example of text being able to generate more revenue, if it is analyzed.

But there is an important second message here as well. Take a look at what customers are saying about promotions. Customers are saying nothing – either good or bad. This means that promotions are having little or no effect on the business. But it is the role of promotions to HAVE an effect on the business. So there is this message to management. Whatever you are doing for promotions is not effective. You need to be doing something else.

Delivering the message that management is not charging enough and that the promotions are ineffective is a VERY important message for management. And these important messages came directly from capturing and analyzing text.

Yet each month management throws these comments away without even trying to understand what the customer is saying.

Last year the restaurant chain announced that is closing a number of its restaurants across the US.


Bill Inmon

Bill Inmon is best-known as the “Father of Data Warehousing” and textual data integration. He has become the most prolific and well-known author worldwide in the data warehousing and business intelligence arena, and has opened the field of textual data integration. In addition to authoring more than 50 books and 650 articles, Bill lectures on data warehousing, textual data integration and related topics. Bill consults with a large number of Fortune 1000 clients, and supports IT executives on data warehousing, business intelligence, and database management issues around the world.

© Since 1997 to the present – Enterprise Warehousing Solutions, Inc. (EWSolutions). All Rights Reserved

Subscribe To DMU

Be the first to hear about articles, tips, and opportunities for improving your data management career.