customer satisfaction

Before

• Call center conversation data is only stored in a database, and it is difficult to know the content of troubles and customer dissatisfaction from conversation data.

After

• By analyzing a huge amount of conversation data, it is possible to visualize customer dissatisfaction and emotions that were not noticed before.

data

Before

• Data divided by organizational department, such as data from development and production departments and data from sales, core systems, and value chain departments, became dark data because it could not be utilized.

After

• Leveraging Splunk's feature of not being influenced by data structure, data from each department is aggregated into Splunk and used as a virtual knowledge base that can be analyzed by each organization to create value from dark data.

Yokogawa Electric Corporation Corporation
IA System & Service Business Headquarters
Life Cycle Service Division
Division Manager Mr. Hiroshi Tanoguchi

Yokogawa Electric Corporation Corporation
IA System & Service Business Headquarters
Life Cycle Service Division
Response Center Management Department
manager Mr. Mr. Sai

Yokogawa Electric Corporation Corporation
IA System & Service Business Headquarters
Product sales center
Business Promotion Section Strategic Planning Gr.
Mr. Shinnya Kato

Yokogawa Electric Corporation Corporation
Digital Strategy Headquarters
Global Application Department
Business Data Utilization Promotion Gr.
manager Mr. Atsuo Kato

Yokogawa Electric Corporation Corporation
Digital Strategy Headquarters
Global Application Department
Business Data Utilization Promotion Gr.
Junko Yamashita

Yokogawa Electric Corporation Corporation
Digital Strategy Headquarters
Information System Department
Business Data Utilization Promotion Gr.
Keita Miyamoto

Visualize levels of satisfaction, desires, and dissatisfaction by analyzing keywords included in conversations

Yokogawa Electric Corporation Corporation (hereinafter referred to as Yokogawa Electric Corporation) has formulated a medium-term plan "Transformation 2020" starting in FY2018. We are trying to strengthen strategic growth investment in digital transformation (DX) as a foundation for working on the three transformations of improving profitability, creating new businesses, and improving productivity.

One such initiative is a project to utilize voice computing at Yokogawa Electric Corporation 's Response Center. Splunk Enterprise (hereinafter referred to as Splunk), provided and supported by Macnica, is used for the core analysis system based on text mining of telephone voices, and production began in November 2018.

The Response Center is a one-stop contact point for call center services that accepts inquiries from customers 24 hours a day, 365 days a year. A "Global Response Center" is set up at the head office in Japan, where dedicated call operators and engineers with deep knowledge of products and services share information in real time. By centrally managing call histories in a dedicated database, we are able to grasp issues while linking them with customer information, maintenance contract details, past histories, etc., and realize prompt and accurate responses. In addition, overseas, which accounts for 70% of our sales, we have set up 9 "Regional Response Centers" to provide the same service and support as in Japan. We strongly back up our service network that spans 230 locations in 80 countries.
“The response center functions as a touch point for customers, and a huge amount of voice data is accumulated while responding to inquiries and problems. Therefore, by analyzing and visualizing conversational voice, which has not been used much until now, the organization can understand the customer's situation and emotions that were not noticed until now. I thought that doing so would be an important key.” (Mr. Tanoguchi)

Although the phone calls sent to the response center were saved, they did not listen to them again except when there was a major complaint. "By extracting keywords from conversations and analyzing them with Splunk, we can numerically visualize customer satisfaction, content of requests, and level of dissatisfaction that we hadn't noticed in the past, leading to subsequent serious accidents and recalls. In addition to being able to extract factors with high precision and efficiency, we hypothesized that it would lead to quality improvement of existing products and the development of new services and products.” (Mr. Yamashita)

Analysis of voice waveforms and keywords in voice text data with Splunk Improvement of customer response by sentiment analysis is also possible

Yokogawa Electric Corporation has been promoting the use of Splunk in-house for several years and boasts one of the largest number of cases in Japan. In 2017, he built a Splunk integration platform and prepared a mechanism that allows a wide variety of data to be linked on the dashboard, and had a wealth of knowledge.

In January 2018, we proposed a plan based on a hypothesis. From there, with the company's unique sense of speed, we made a budget in April and conducted PoC (proof of concept) over a month until May. The PoC also incorporated the latest features of Amazon Web Services (AWS). Specifically, the contact point of the response center is built with the cloud-based contact center "Amazon Connect", and conversation records, operator logs, and CTI data are collected. When you put them into the "Amazon S3" bucket, Splunk automatically converts the voice into text using the speech recognition service "Amazon Transcribe", and performs keyword detection and emotional element waveform analysis processing with "Amazon Comprehend". . By importing it into Splunk together with customer contract data, past CTI data, service reports, etc., natural language processing (NLP) analysis is performed to detect insights and relationships, and visualized on various dashboards. That's the flow. All of these are event-driven and serverless with AWS Lambda.

As a result of the PoC, we were able to confirm the possibility of improving the responsiveness of the response center and improving customer satisfaction.In addition, we were able to visualize the end users who are using old products, which provided an opportunity to collect information directly without going through an agency. It was expected that it would be possible to escalate to sales.

Particular attention was paid to the improvement in accuracy by including data from Empath, an engine that analyzes emotions from speech waveforms provided by Empath, which develops and sells speech emotion analysis AI. At the beginning of the demonstration experiment, if there were no negative keywords, the analysis results were recognized as "normal", but by incorporating voice waveform analysis into Splunk from the middle, it became clear that what was normal was actually angry.
"It was a big surprise when we saw that on the dashboard, because in production, if we detect that a customer might be angry, it's better to escalate it to a sales rep quickly instead of trying to address it at the call operator stage, so we can take more effective action quickly. I think it is very effective to change the behavior according to the customer's emotions Shinnya Kato."

In addition, by utilizing Splunk, it has become possible to solve the problem of information silos in the manufacturing industry. In the internal information flow, the flow of product/service development axis information in development and production is separated from the customer management axis information flow in sales. The database was fragmented, including the flow of information along the maintenance value chain axis, making it impossible to visualize in a coordinated manner.
“It is not easy to aggregate the knowledge that each business unit or department has built and accumulated for its own purpose into a database. data can be easily imported.
By visualizing the data of each department and making it available for their respective purposes, we expect that it will be used as a virtual knowledge base. In the future, I believe that we will be able to use after-sales service information for product planning, and to feed back information on how to respond to inquiries to production preparation." (Mr. Choi)

Starting speech data analysis from English-speaking countries where NLP is advantageous

Response Center speech analysis by Splunk moved to the production stage in November 2018. For the time being, we have started analysis using actual voice data at the regional response center in North America. In the future, we plan to put the operation on track while accumulating data, and gradually expand it to the European region other than North America. “Currently, we are operating based on English, which is relatively easy to use in NLP, but we will improve the accuracy by promoting multiple analyzes such as incorporating voice waveform analysis data, and we will be able to use it in various languages in the future. There are still many things to do, but someday I would like to work on Japanese, which is said to be a disadvantage for NLP.” (Mr. Tanoguchi)

In the future, we will be able to expand the range of data to be imported, such as CRM data and email data, and link AI and deep learning technology with Splunk, enabling us to provide flexible responses suitable for customers from the analysis of large amounts of data. It is said that we will continue to do so.
“Because Splunk is a data billing system, it can be used freely by many people, and the processing speed is fast, and the secure operation that can limit the logs to be imported was evaluated as a point.This project could not have been realized without Splunk. Sho.” (Mr. Miyamoto)

Yokogawa Electric Corporation 's recent example of combining Amazon Connect, Amazon Comprehend, etc., converting call center speech into text and analyzing it with Splunk is extremely pioneering not only in the manufacturing industry, but is attracting attention from around the world.
“Although there were concerns that speech analysis was premature, the Company believe that the excellent output produced by Splunk as a core is a result that expands future possibilities. I feel there is great potential in this." (Atsuo Kato)

User Profile

045-476-2010
月～金 8：45～17：30

inquiry
Please feel free to contact us

Free seminar
Click here for various seminars

Document download
Click here for various materials