🔬 Conversation Analysis Guide

This guide explains how to use Uspeech Analytics to extract meaningful insights from interview transcripts and conversation data.

---

Overview

Conversation Analysis transforms your interview transcripts into structured, actionable insights. Using AI-powered analysis, you can automatically extract:

• Pain Points — Problems, frustrations, and challenges mentioned by participants
• Gains — Desired outcomes, benefits, and positive experiences
• Jobs to Be Done — Tasks, goals, and objectives participants are trying to accomplish
• Forces of Progress — Motivations and barriers affecting behavior change
• Affinity Maps — Thematic groupings of related insights
• Custom Questions — Answers to specific questions you define

This is invaluable for UX research, customer discovery, product development, and understanding user needs at scale.

---

Quick How-To

How to Run Conversation Analysis Using Predefined Methodologies

1. Prepare your transcripts

   • Transcribe your audio files first (see Transcriptions Guide)
   • Or upload existing text transcripts

2. Select project for analysis

   • Click Conversation Analysis on the navigation bar
   • Choose your project from the dropdown box

3. Choose analysis type

   • Click Add Analysis and select from the available analysis types
   • Choose based on what insights you need (see Analysis Types below)
   • Check Add Citations if you want to include direct quotes from participants in the analysis
   • (Optional) Put additional context or instructions in the Additional Context field
   • Click Add to start the analysis

How to Run Custom Conversation Analysis

1. Prepare your transcripts

   • Transcribe your audio files first (see Transcriptions Guide)
   • Or upload existing text transcripts

2. Select project for analysis

   • Click Conversation Analysis on the navigation bar
   • Choose your project from the dropdown box

3. Choose custom analysis type

   • Click Custom questions
   • Choose either Enter questions or From template
   • If Enter questions is chosen:
     • Enter your questions in the text area. You can add as many questions as needed.
   • If From template is chosen:
     • Select a template from the dropdown (see Templates & Codes Guide)
   • Optionally, you can choose “Analyze separately” to get answers from each conversation individually
   • Click Add to start the analysis

How to Review the Results

1. Review results
   • Analysis typically takes 1-5 minutes depending on transcript length
   • Once complete, the status of the corresponding analysis type will change to “Completed”
   • Click on the analysis type to view the results
   • Results show categorized insights with supporting quotes
   • For “Custom questions” analysis, you can see the answers to your specific questions
     • If “Analyze separately” was chosen, you’ll see multiple tabs with answers on each question for each conversation, with supporting quotes
     • If “Analyze separately” was chosen, you’ll also see a codification tab showing stats for the main answer categories across all conversations.

💡​ Tip: You’ll get an email notification when the analysis process is finished. In order to enable email notifications, go to Profile, check “Notify job completion” and click Save.

2. Export findings
   • Check one or more analysis types to export
   • Click Export and choose between Word document or Excel file

---

Analysis Types

Pain Points Extraction

What it finds: Problems, frustrations, challenges, and negative experiences mentioned in conversations.

Best for:

• Identifying product improvement opportunities
• Understanding customer complaints
• Prioritizing issues to address

Output includes:

• Categorized pain points with descriptions
• Direct quotes from participants supporting each pain
• Frequency indicators showing how common each pain is

Example output:

Pain Category	Description	Supporting Quotes
Long wait times	Users frustrated by delays in customer service	”I waited 45 minutes just to speak to someone…”
Confusing navigation	Difficulty finding features in the interface	”I couldn’t figure out where the settings were…”

---

Gains Extraction

What it finds: Desired outcomes, benefits sought, and positive experiences participants describe.

Best for:

• Understanding what users value
• Identifying feature opportunities
• Crafting value propositions

Output includes:

• Categorized gains with descriptions
• Supporting quotes from participants
• Insights into user motivations

Example output:

Gain Category	Description	Supporting Quotes
Time savings	Users want to complete tasks faster	”If I could do this in half the time, that would be amazing…”
Peace of mind	Users value reliability and trust	”I just want to know it’s going to work every time…”

---

Jobs to Be Done (JTBD)

What it finds: The functional, emotional, and social tasks participants are trying to accomplish.

Best for:

• Product strategy and positioning
• Understanding user motivations
• Innovation and new feature development

Output includes:

• Jobs categorized by type (functional, emotional, social)
• Job statements in standard JTBD format
• Supporting evidence from interviews

Job Types:

Type	Description	Example
Functional	Practical tasks to complete	”Get to work on time”
Emotional	How users want to feel	”Feel confident in my decisions”
Social	How users want to be perceived	”Appear professional to colleagues”

---

Forces of Progress

What it finds: The four forces that drive or prevent behavior change, based on Jobs-to-be-Done methodology.

Best for:

• Understanding adoption barriers
• Crafting persuasive messaging
• Reducing friction in user journeys

The Four Forces:

Force	Description	Questions It Answers
Push	Circumstances making users unhappy with current situation	”What’s wrong with how things are now?”
Pull	Attraction to a better future state	”What’s drawing them toward change?”
Anxiety	Fears and uncertainties about change	”What’s holding them back?”
Habit	Existing behaviors that resist change	”What routines would they have to break?”

Example output:

• Push: “The old system crashes constantly” — frustration driving change
• Pull: “I’ve heard the new version is much faster” — appeal of solution
• Anxiety: “What if I lose all my data during migration?” — adoption barrier
• Habit: “I’ve used this workflow for 5 years” — resistance to change

---

Affinity Mapping

What it finds: Themes and patterns across multiple interviews, grouped by similarity.

Best for:

• Synthesizing large amounts of qualitative data
• Identifying common themes across participants
• Preparing data for further analysis

Output includes:

• Thematic clusters of related insights
• Key quotes organized by theme
• Visual hierarchy of findings

---

Custom Questions

What it finds: Answers to specific questions you define.

Best for:

• Targeted research questions
• Following up on specific hypotheses
• Extracting particular types of information

How to use:

1. Select “Custom Questions” analysis type
2. Enter your questions (one per line)
3. The AI will search transcripts for relevant answers

Example questions:

• “What features do participants mention wanting?”
• “How do participants describe their onboarding experience?”
• “What competitors do participants mention?”

Bulk analysis and per-conversation analysis:

• When analyzing multiple files, you can choose to analyze them together or separately
• When analyzing separately, you will get answers to your questions for each conversation individually. In addition, you will get most common categories of answers on each question across all conversations, similar to Survey Analysis
• When analyzing together, you will get answers to your questions across all conversations

---

Analyzing Multiple Files

Bulk Analysis

Conversation analysis is project-wide by default. All transcribed conversations in your project are analyzed together to identify themes and patterns.

Audio files and their transcriptions are organized into projects based on a shared topic, study, or set of questions. You control which files belong to the same project, so make sure you add conversations that you want analyzed together.

Benefits of bulk analysis:

• Identifies patterns across multiple participants
• Automatically groups similar insights
• Shows frequency of themes across interviews

Best Practices for Bulk Analysis

• Similar content: Keep each project focused on one study, topic, or research phase
• Reasonable quantities: Projects with 5-20 transcripts work well; very large projects may need to be split, if the audio type is not call center calls

---

Understanding Results

Result Structure

Analysis results typically include:

1. Category/Theme Name — A descriptive label for the insight
2. Description — Brief explanation of what this category represents
3. Supporting Quotes — Direct quotes from participants that support this insight
4. Frequency — How many interviews/participants mentioned this (for Custom Questions analysis if analysis is done per conversation)

Working with Quotes

Quotes are extracted directly from your transcripts:

• They provide evidence for the insights
• Use them in reports and presentations
• They help stakeholders understand user voice

Exporting Results

Word Document (.docx):

• Formatted report with all insights
• Easy to share and edit
• Good for including in research reports

Excel (.xlsx):

• Structured data in spreadsheet format
• Good for further analysis
• Easy to filter and sort

---

Tips for Best Results

Preparing Transcripts (if you load them manually)

1. Quality matters — Accurate transcripts lead to better analysis
2. Include context — Full conversations work better than excerpts
3. Review speaker labels — Correct any obvious errors before analysis
4. Transcripts should be in SRT format — This is the only format supported for manual upload

Choosing Analysis Types

Research Goal	Recommended Analysis
Understand user problems	Pain Points
Identify opportunities	Gains
Product strategy	Jobs to Be Done
Understand adoption	Forces of Progress
Synthesize many interviews	Affinity Mapping
Specific questions	Custom Questions

Iterating on Analysis

• Start with Affinity Mapping for a broad overview of themes
• Follow up with Pain Points or Gains for specifics
• Use Custom Questions to dig into particular areas

---

Frequently Asked Questions

Q: How long does analysis take? A: Typically 1-5 minutes per file, depending on transcript length and analysis type.

Q: Can I analyze the same file multiple times? A: Yes, you can run different analysis types on the same transcript. This can be done simultaneously or sequentially.

Q: How accurate is the analysis? A: The AI extracts insights based on what’s explicitly stated in transcripts. Always review results to ensure they match your interpretation.

Q: What if important insights are missed? A: Try running Custom Questions analysis with specific prompts, or review transcripts manually for nuanced insights.

Q: Can I edit or add to the analysis? A: Export to Word or Excel format to add your own annotations and insights.

Q: How many files can I analyze at once? A: We recommend batches of 5-20 files for optimal results. Very large batches can be split. Usually, contact center calls are shorter and can be analyzed in larger batches.

---

Next Steps

• Learn about Survey Analysis for coding open-ended responses
• Create Templates to standardize your analysis process
• Return to Quick Start Guide for an overview