SayPro Guide: High-Volume Thematic Analysis Using GPT Prompts

Goal: Extract and analyze up to 100 distinct themes/topics per prompt from qualitative data using SayPro-approved GPT prompt structures.

---

1. Purpose

To process large volumes of stakeholder feedback, open-ended survey responses, or interview transcripts efficiently and consistently using generative AI, while maintaining methodological rigor and alignment with SayPro’s reporting standards.

---

2. Requirements

✅ Inputs Needed

• Cleaned qualitative data (survey comments, focus group transcripts, etc.)
• Approved SayPro GPT prompt templates (see below)
• Categorization rubric (optional, e.g., Education, Employment, Health)

✅ Output Requirements

• Up to 100 unique, relevant, and non-redundant themes
• Frequency count or relevance rating per theme
• Grouped under SayPro programmatic areas (if needed)

---

3. Approved SayPro GPT Prompt Template for Thematic Extraction

🔹 Prompt Template: Extract 100 Themes

pythonCopyEditYou are an expert evaluator for SayPro. From the following text, extract and list up to 100 unique, distinct themes or topics mentioned. For each theme, include:
1. Theme Title (4-6 words max)
2. Short Description (1-2 sentences)
3. Frequency (number of mentions)
4. Associated Keywords or Sample Phrases

Text:
"""
[Paste all qualitative feedback or interview content here]
"""
Output format:
- Theme #
- Title:
- Description:
- Frequency:
- Keywords:

---

4. Optional: Group Themes by Domain

You may request GPT to organize themes into SayPro domains:

• Youth Empowerment
• Digital Skills Development
• Public Health Education
• Community Engagement
• Monitoring & Feedback
• Policy Awareness
• Access & Equity

📌 Modified Prompt Example:

pgsqlCopyEditGroup the extracted themes into SayPro’s programmatic domains. For each domain, list up to 15–20 themes. Provide counts and sample keywords under each theme.

---

5. Post-Processing and Analysis

After receiving the GPT-generated themes:

📊 Analysis Steps:

• Import into Excel or Google Sheets
• De-duplicate and standardize phrasing
• Assign thematic weights or relevance scores
• Use pivot tables to group by domain, location, or sentiment
• Visualize top themes using word clouds, bar charts, or heatmaps

---

6. Sample Output Snippet (Example)

Theme #	Title	Description	Frequency	Keywords
01	Access to Internet	Many respondents cited unstable or no access to internet as a major barrier to digital training.	34	“no WiFi”, “signal issues”, “data expensive”
02	Gender-Sensitive Programming	Women requested more safe spaces and gender-specific job training modules.	18	“women-only”, “safe space”, “gender programs”
03	Delays in Stipends	Delays in receiving stipends or disbursements affected training attendance.	21	“late payment”, “stipend missing”, “transport refund”

---

7. Scaling for Large Batches

If analyzing multiple datasets:

Dataset	Prompt Batch #	GPT Output File
Northern Region Survey	Batch 01	northern_q2_themes.csv
Youth Voice Interviews	Batch 02	youth_voices_analysis.xlsx
Feedback Box Comments	Batch 03	comments_theme_report.docx

---

8. Recommended Tools for Managing Themes

• Google Sheets / Excel – For sorting, tagging, frequency analysis
• ChatGPT or API Access – For batch prompt submission
• Power BI / Data Studio – For visualization of top 100 themes
• SayPro Prompt Tracking Sheet – To log prompt versions and output files

---

9. Internal QA Checklist (Based on SayPro QA Standards)

Check	Pass/Fail	Notes
Were 100 unique, meaningful themes extracted?	✅
Were any duplicates or overlaps identified?	❌	Theme 14/15 are similar
Were outputs sorted by relevance or frequency?	✅
Was approved GPT prompt used without modification?	✅
Have outputs been reviewed by M&E officer?	✅	Approved on June 2

---

✅ Need a Live Template or Prompt Sheet?

I can provide:

• A pre-formatted Excel template to track 100 themes
• A fillable GPT prompt sheet for staff to use
• An automated script for API-based thematic extraction