Survey Analyzer
Extract themes from thousands of responses
Process open-ended survey responses at scale to identify themes, sentiment patterns, and quantitative insight clusters. Generates executive summaries and per-segment breakdowns that manual thematic analysis would take weeks to produce.
GPT-6 Luna
for 800K tokens a month · 80% input / 20% output
Why this model
For this workload — 800K tokens a month, 80% of them input — it comes to about $0.14 a month, the lowest of the models suggested here. Its 1M-token context window holds long inputs in a single call, which matters for this kind of work.
Alternative models
Implementation tips
- 1
Use a two-stage pipeline: first ask the model to assign each response to one of 10–15 themes (cheap, fast), then pass the grouped themes to a more powerful model for nuanced synthesis — this hybrid approach cuts cost by 70% vs. analyzing each response deeply.
- 2
Process responses in batches of 50 and ask for consistent theme labels — define the theme vocabulary in your system prompt and instruct the model to use only those labels to ensure responses are comparable across batches.
- 3
Include demographic metadata alongside each response in your prompt — the model can surface 'junior employees feel X but senior employees feel Y' patterns that aggregate analysis misses.