Analysis Workflow

This toolkit provides an end-to-end pipeline for processing survey data, generating mock LLM responses, and evaluating those responses against real data.

1

Convert TSV to JSON

Upload your TSV survey data file and convert it into structured JSON files - one file per respondent.

TSV Parsing JSON Creation
2

Infer Consumer Attributes

Process each JSON file to infer attributes about consumers using LLM analysis.

Consumer Profiling Attribute Extraction
3

Generate Mock Responses

Use LLMs to generate mock responses based on the inferred consumer profiles.

Response Simulation Batch Processing
4

Evaluate Mock Responses

Compare generated mock responses against actual responses to evaluate accuracy.

Response Comparison Accuracy Scoring
5

Visualize Results

See a visualization of the evaluation results to gain insights into LLM performance.

Interactive Visualization Performance Metrics

1Data Processing Tools

TSV to JSON Converter

Upload a TSV file containing survey responses and convert it into individual JSON files for each respondent, packaged as a ZIP archive.

Open Tool
Consumer Attribute Inferencer

Upload the ZIP archive of JSON files and use LLM calls to infer attributes for each consumer profile, creating enriched data.

Open Tool

2Response Generation & Evaluation

Mock Response Generator

Use the enriched consumer profiles to generate realistic mock responses via LLM, simulating how consumers might answer specific questions.

Open Tool
Response Evaluator

Compare the generated mock responses against real answers to evaluate accuracy, reasoning, pattern recognition, and insight quality.

Open Tool

3Results Analysis

Evaluation Visualizer

Visualize the evaluation results with interactive charts and detailed breakdowns of how different LLMs performed in predicting consumer responses.

Open Tool