Traditional Insights, Modern Approaches: 6 AI-assisted Toolkits for Qualitative Researchers Milind Kelkar – Chief Analytics Officer, Unpickle

In the evolving landscape of qualitative data analysis, the role of AI is not to replace the human element of analysis, but rather to enhance it. At the heart of our approach lies the belief that technology serves best as an assistant, easing the more laborious aspects of data handling and allowing researchers to delve deeper into the parts of data that truly resonate with their research objectives. This shift of focus from tedious manual data coding to engaging with the essence of the qualitative data signifies more than just a procedural change; it represents a return to the core of qualitative research — a deep, empathetic understanding of the human narrative. Discover how our suite of 6 AI-assisted toolkits is redefining this journey, inviting researchers to explore a harmonious blend of traditional insights and modern technological approaches.

Background

Our project commenced a year ago, coinciding with the launch of Large Language Models like OpenAI's ChatGPT, Claude, Mistral, and Cohere. This convergence signaled a pivotal shift in Qualitative Data Analysis capabilities, heralding a new era in the field. It compelled us to re-evaluate traditional methodologies and embrace more advanced, AI-driven approaches. We're now at a juncture where we can articulate the practical implications of these technologies. Our development teams have been diligently working to transform these emerging concepts into actionable, cutting-edge toolkits, marking a significant stride in qualitative research.

Why

These toolkits offer the ability to validate codes while Researchers are immersed in reading transcripts, shifting focus from the manual aspects of data management to the more customer focused tasks of synthesis and storytelling. They allow for a 360-degree view of qualitative data, enabling researchers to handle multiple transcripts and large, diverse data sources simultaneously — from customer-shared photos to social media narratives — enriching the insights with a broader range of perspectives. They support various coding styles, adapting to the nuances of each research project and the evolving needs of clients.

 These toolkits are not just about keeping pace with technology; it's about augmenting Researcher’s capabilities, allowing them to connect more deeply with the customer stories they seek to understand.

Toolkit 1a: Data Sorting/Content Analysis

Description: In qualitative research, Data Sorting plays a crucial role in organizing diverse non-numeric data, such as interviews, focus groups, customer feedback, diary entries, moderator notes, and social media responses. It serves to discern patterns and relationships, effectively aligning direct (verbatim) responses and associated data with pertinent analysis questions or themes.

When to use: Data Sorting excels in transforming unstructured data into a structured format, a process that greatly benefits researchers. This toolkit streamlines the analysis by segmenting extensive questionnaires into smaller, manageable parts, allowing for more thorough and focused examination, especially when addressing specific research questions.

Action: Researchers can interactively click on any cell (for instance, 'Amount & Stationary Products') to access verbatim responses or auto-summarized outputs from various interviews, streamlining data review and analysis.

 Table: Data Sorting in Excel

 Table: Data Sorting output for Amount spent on Stationary product for Respondent 1 and 2

Toolkit 1b: Automated Summarization

Description: Automated Summarisation distill key information from large volumes of qualitative data. This tool efficiently condenses verbatims into succinct summaries. By extracting the essence of extensive data, it provides researchers with an immediate understanding of overarching themes and insights without the need to comb through every detail manually.

When to use: This toolkit is particularly useful when researchers are faced with tight deadlines or vast amounts of qualitative data. It is ideal for initial reviews, where a quick understanding of the data's main points is necessary, or when needing to present preliminary findings to stakeholders without delay.

Action: The table provided demonstrates a side-by-side comparison: on one side, you'll find the raw verbatim responses, and on the other, the Auto Summarised outputs complemented with representative quotes. This comparison illustrates the efficiency and effectiveness of the summarisation process.

 Table: Auto Summary output for Drivers of current preferred pen

Toolkit 2a: Qualitative Coding - Line by Line Qualitative Coding in MS Excel

Description: Toolkit 2a offers a unique hybrid approach to qualitative data analysis, merging the efficiency of AI-powered coding with the nuance and insight of expert human validation within MS Excel. This innovative process enables researchers to swiftly identify themes or patterns and drastically shortens the timeline for insight extraction.

When to use: Ideal when researchers are pressed for time yet require a comprehensive understanding of coded data without the labor-intensive task of examining entire transcripts. It excels in situations demanding cohort wise analysis, offering nuanced insights specific to different groups or categories. Additionally, its adaptability makes it suitable for various research methodologies, including grounded theory, content analysis, thematic, narrative, and discourse analysis, thereby broadening its applicability.

Action: Researchers are provided with an all-inclusive coded dataset complemented by an in-depth spreadsheet. This spreadsheet meticulously details each Code, the corresponding Quote, the Tonality of the responses, and a Code Description.

 Table: Line by Line Qualitative Coding in Excel 

Toolkit 2b: Qualitative Coding – Coding in MS Word

Description: Toolkit 2b streamlines the coding process of qualitative research by leveraging the familiar environment of MS Word. This tool allows for intuitive interaction with interview transcripts, using Word’s comment feature for coding. Researchers can effortlessly add, modify, or remove codes at any stage of their analysis, providing flexibility and ease of use.

When to use: This approach is particularly useful for those who prefer a more hands-on method of engaging with their data or for working with legacy qualitative data analysis platforms.

Action: Researchers can import the MS Word documents with embedded codes into various QDA platforms, facilitating further in-depth analysis and interpretation of the data.

Table: Qualitative Coding in MS Word

Toolkit 3: Advanced Memoing Techniques

Description: Toolkit 3 introduces Advanced Memoing Techniques, a powerful tool designed to deepen the qualitative data analysis process. Memoing enables researchers to thoroughly understand their data by uncovering and examining patterns, themes, concepts, and relationships within it. This process not only aids in the interpretation of data but also allows researchers to reflect on the data, the research process itself, and their personal reactions and thoughts about the data. This reflective practice is instrumental in developing a richer, more nuanced understanding of the qualitative research.

When to use: Memoing is particularly effective when used in conjunction with coding. As researchers assign codes to data segments, memos serve as an invaluable tool to record the rationale behind code assignments and to explore the connections between different codes. This approach is especially useful for researchers who wish to provide a comprehensive account of their coding decisions and the interrelationships within their data.

Action: The analysis should also consider looking for patterns across other responses to determine if there are additional corroborating experiences, or if this feedback is an outlier. It would be crucial to investigate both the frequency and context of similar issues to formulate an effective response strategy for the company.

Table: Memo for Respondent 1 on issues faced with pen 

Toolkit 4: Transformation of Codes into Deductive/Inductive Categories

Description: Codes need to be converted into categories and themes. Categories could originate basis experience or based on Grounded theory or product under consideration.

When to use: Categories and themes help to further summarize data. Its relevant when Researcher has large number of codes that needs to be grouped.

Action: Codes have been converted into Deductive codes

Toolkit 5: Drivers of business outcome in Mixed Methods and Surveys

Description: Codes and tonality serve as explanatory variables to understand various business outcomes, such as purchase intention, consideration set, likeability, and showroom visits. These variables are crucial for connecting qualitative insights with quantitative metrics.

When to use: This toolkit is particularly useful for large qualitative studies, including those involving product placements, spontaneous customer feedback, and open-ended verbatim responses in surveys. It can help uncover the factors that influence the mentioned outcomes.

Action: Analyze qualitative data using a coding system, focusing on themes or categories. Calculate the volume of code mentions and assign importance scores to understand their impact on outcome variables. Integrate coded insights with quantitative data, and report findings to explain the drivers of business outcomes.

Toolkit 6: Dynamic Image Coding

Description: Toolkit 6 codes images, enabling researchers to annotate, categorize, and analyze visual data with the same rigor as textual data. It facilitates the identification of patterns, themes, and insights within visual content, enhancing the depth and breadth of qualitative analysis.

When to use: This is useful for researchers dealing with large volumes of visual data, such as photographs, videos, or other graphic materials. It's ideal for projects where visual elements play a key role in conveying meaning, such as in ethnographic studies, media analysis, or user experience research.

Action: Analyze image-coded data for specific applications, such as calculating the frequency of occurrences in studies like beverage consumption patterns. Moreover, it can integrate with text-coded data, enabling researchers to draw comprehensive conclusions that combine both textual and visual insights.