How to use generative AI to translate survey analysis into next actions

Challenges related to surveys and customer feedback

First, let's examine the current situation regarding why surveys and customer feedback (VoC: Voice of Customer) within companies are often left unused.

The current state of scattered and underutilized data

Companies possess a wide variety of data sources, including survey data collected via the internet, paper-based data, and call center voice logs. Because these sources are in different formats, it is difficult to manage them all together, and as a result, it is not uncommon for some data to be left unused.

While digitized data, such as web surveys, is relatively easy to handle, manual data entry for paper surveys or transcription of call center recordings is far more time-consuming. When multiple people compile data in different formats, it's easy to end up in a situation where consistent analysis becomes impossible.

Manually aggregating this disparate data is time-consuming and costly, inevitably leading to in-depth analysis of free-response text and voice data being pushed to the back burner. As a result, opportunities to leverage insights from customer feedback are missed.

Processing load of free-form text and audio data

While multiple-choice responses can be automatically tallied using data aggregation tools, free-response answers and audio data require additional work such as transcription and meaning extraction. Especially with large amounts of data, manual review becomes impossible, resulting in the data being less effectively utilized.

Understanding the thoughts and feelings expressed in the open-ended responses requires reading and categorizing each one individually. This process, including transcribing audio logs, is even more labor-intensive. As a result, some organizations lack the resources to handle such tasks, ultimately leaving open-ended comments and call recordings largely unused.

Back to Top

Digitization of analog data and text extraction

As preparation for utilizing the raw data held by these companies, we will introduce the steps to convert unstructured data such as paper documents and audio recordings into a digital format that can be treated as text.

Digitalization of paper questionnaires

When digitizing questionnaires collected on paper, the first step is to capture them as PDFs or images using a scanner or smartphone, and then extract the text data using character recognition technology.

Speaking of character recognition technology, traditional OCR excelled at recognizing printed characters, but recent multimodal LLM (generative AI that supports various file formats) can flexibly handle handwritten characters as well.

Specifically, we will design a workflow where scanned image files are uploaded (requested) to a multimodal LLM, the results are converted to text, and then reviewed and corrected. This is expected to significantly reduce working time compared to manual data entry by humans.

This digitized data can be handled in a consistent format with data obtained from other channels. As a result, it becomes easier to compare and aggregate the data. Furthermore, by tagging the information extracted during the digitization process, subsequent analysis processes can be made even smoother.

▼I want to know more about generative AI
⇒ Generative AI | Glossary

Text conversion of audio logs

The process of converting audio data, such as recordings of phone conversations or face-to-face interviews with customers, into text is also important. Similarly, text can be extracted from audio using a multimodal LLM (Language-Limited Management) that has a speech-to-text conversion mechanism.

First, the call is transcribed and the results are saved as a text file. After that, it is possible to use generative AI or natural language processing tools to tag each speaker or classify the content of the speech according to specific criteria.

By processing this diverse raw data into text, it becomes possible to arrange the data in the same format as the survey results and analyze it from multiple perspectives.

Back to Top

Data classification and structuring using generative AI

Simply extracting text yields limited insights. Here, we introduce a method for tagging and classifying text data using generative AI, thereby structuring it in a meaningful way.

Automatic determination of attributes and categories

Because our generative AI has learned from a vast amount of natural language, it can automatically identify products, inquiry types, sentiments, and other factors from open-ended text responses and assign them as tags. This allows us to get an overview of massive amounts of data at once and instantly understand which categories of opinions are most prevalent.

Furthermore, with careful consideration of the prompts, even detailed classifications are possible. By tagging along multiple axes, such as whether the opinion is categorized by product category or region, it becomes easier to analyze the data from various perspectives later on.

The same applies to sentiment analysis; by giving instructions to the generating AI to categorize responses as either positive or negative, it becomes clear which opinions should be prioritized for follow-up. Combining these approaches allows customer service and support departments to implement improvement measures more efficiently.

Extracting urgency and insights

Not all opinions carry the same weight. Some opinions contain strong alert signals, such as "an urgent response is needed." Generating AI can determine this level of urgency based on keywords and the overall flow of the text.

Traditional methods such as survey analysis and manual review by staff made it difficult to screen large amounts of data in a short period. By utilizing generation AI, it becomes possible to detect urgent words and negative nuances contained in individual sentences and set priorities accordingly.

At the same time, we can sometimes decipher customers' latent needs and unexpressed dissatisfactions from the wording of their texts. As a result, we can not only prevent delays in handling complaints, but also identify potential areas for service improvement and hints for new products at an early stage. This directly leads to improved customer satisfaction and enhanced brand image, significantly improving overall operational efficiency and the quality of customer service.

Back to Top

Utilizing and visualizing organized data

Data generated by AI can be easily analyzed from multiple perspectives, both quantitatively and qualitatively, and visualized using BI dashboards and other tools.

Understanding the overall trend and trend analysis

By arranging multi-faceted tagged data chronologically, you can see at a glance how customer feedback is changing. For example, by looking at the trend of complaints that arose after the launch of a new product, you can quickly identify areas that need improvement.

Furthermore, visualizing emotional shifts and the frequency of specific keywords allows us to understand market trends and changes in customer satisfaction, providing valuable information for considering when to implement countermeasures. Sharing the analysis results with the entire team also has the advantage of making it easier to implement consistent measures.

Organizing information in a visually understandable format allows for an objective assessment of which issues are the top priorities for customers. In particular, addressing issues where negative feedback is concentrated early can yield significant results.

Furthermore, by delving deeper into product and service areas that have received a large number of positive comments, it is possible to reaffirm strengths that differentiate you from competitors and utilize them in new marketing strategies.

Faster individual response and improved service

Automating the selection of high-priority feedback makes it easier to respond quickly to inquiries from important customers. Visualizing increases in inquiries related to specific products allows for updates to response manuals and adjustments to staffing levels. Understanding customer trends in real time enables service improvements that keep you ahead of the competition.

By creating a system that filters data based on emotional analysis using AI generation and notifies only of negative and urgent cases, the speed of complaint handling can be dramatically increased.

By categorizing inquiries by department, it becomes easier to route inquiries requiring specialized knowledge to the appropriate department immediately.

Back to Top

data integration and data preparation using AI-generated data.

For companies to continuously gather customer feedback, automating the data collection and analysis process as much as possible is key. This section explains the concepts for building a sustainable workflow that automatically connects dispersed data from collection to analysis.

Automated integration flow from data collection to analysis

First, we clearly define the data collection point and design a workflow that links it to various tools, including AI-generated data. It is important to design the system so that once paper questionnaires are scanned or audio recordings are digitized, classification and tabulation are performed automatically, and the data is imported into a database for analysis.

At the data entry point, we provide entry points tailored to each format, such as devices for scanning paper questionnaires or call center call recording systems. From there, we build a system to centrally collect information through APIs, file uploads, and other means.

Subsequently, the data is converted to text and automatically tagged using AI generation, and category and sentiment information is added when the results are written to the database. Finally, by linking the data to visualization tools and reporting systems, it becomes possible to efficiently gain insights.

Creating insights through continuous data management

By continuously operating a system once it has been built, new survey results and customer feedback are constantly being accumulated, and the accuracy of the analysis improves as the amount of data increases.

Automating data preparation frees up staff to develop strategies based on reports and analysis results. Additionally, it reduces inconsistencies in data extraction and classification, resulting in clearer decision-making processes.

By constantly incorporating and analyzing the latest customer data, it becomes possible to detect changes in trends in real time and implement measures at an early stage. When such a data-driven organizational culture takes root, the speed and quality of product development and service improvement will increase dramatically.

Back to Top

Finally

Until now, while the importance of open-ended responses in surveys and customer feedback has been recognized, it has been difficult to fully utilize them due to the enormous amount of time and effort required. However, with the advent of generative AI, an environment is being created where disparate data formats such as paper and audio can be efficiently digitized and compiled into consistent data. This makes it possible to redefine valuable opinions that were previously overlooked as corporate assets.

Data classification and structuring using generative AI goes beyond simply streamlining aggregation tasks. By incorporating additional information such as emotions and urgency into the data, it becomes possible to objectively grasp priority issues and potential needs. As a result, a major benefit is that field personnel can engage in prompt customer support and highly accurate product development based on solid evidence, without relying on subjective opinions.

The key is to ensure these processes are not temporary but become an established, automated "data integration" system. By building a pipeline that seamlessly connects data collection, analysis, and insight extraction, the entire organization can evolve into a data-driven system that constantly senses customer changes and can respond flexibly.

Back to Top

The person who wrote the article

Affiliation: Data Integration Consulting Department, Data & AI Evangelist

Shinnosuke Yamamoto

After joining the company, he worked as a data engineer, designing and developing data infrastructure, primarily for major manufacturing clients. He then became involved in business planning for the standardization of data integration and the introduction of generative AI environments. From April 2023, he will be working as a pre-sales representative, proposing and planning services related to data infrastructure, while also giving lectures at seminars and acting as an evangelist in the "data x generative AI" field. His hobbies are traveling to remote islands and visiting open-air baths.
(Affiliations are as of the time of publication)

Recommended Content

Related Content