Survey Data Analysis: a Tokeny case study
Import survey data for automatic option distribution, cross-analysis, and visual report generation. Follow the recorded task, execution screenshots, actual files and verification notes.
By the Tokeny team · Windows · Tokeny 1.7.8 · gpt-6-sol
Prepare the inputs
Create a separate workspace in Tokeny, download and unpack the sample inputs, then select a model that supports tools. The sample uses synthetic data or the public sources named in the article.
- survey.csv
Send the task to Tokeny
The original task was executed in Chinese. Copying these instructions reproduces the same task; the results below include an English report.
分析 12 份合成问卷，统计满意度和功能偏好分布，按角色做交叉分析，用离线 HTML 展示图表并写结论。 交付文件：survey.html、analysis.csv。 验收要点： - 样本量 12 - 均值 3.75 - 明确小样本不能代表总体 这是使用合成演示输入或指定公开资料的真实操作案例。请先读取工作空间输入，再实际生成文件并自检。所有写入、脚本与临时服务限于当前工作空间；临时服务只监听 127.0.0.1，执行后关闭；不触达外部收件人、不发布、不付款、不更改系统设置，不调用子智能体。需要公开资料时可读取给定 URL。只启用完成任务必需的工具。另写 report.md（中文）和 report.en.md（英文），每份 400–800 字，包含输入、处理步骤、真实结果、检查方式和局限；失败或未验证部分如实记录，不能宣称成功。最终回复用中文列出产物、关键结果与待核实项。
The recorded run used an isolated demonstration workspace. Review the commands and file paths before approving operations in your own workspace.
What actually ran
The task was submitted through the Tokeny input box. An online model read the inputs and executed tools. This run covers the local sample task and its actual files. Results and checks are recorded below.
Results and verification
Input
This analysis reads survey.csv from the current workspace. The file contains 12 synthetic questionnaire records with an ID, role, satisfaction score from 1 to 5, preferred feature, and weekly usage hours. The valid sample size was checked as 12. The three roles are Operations, Development, and Teaching, with four records each. These are synthetic demonstration inputs rather than a real-world probability sample.
Processing
A local Python script parsed the CSV and converted satisfaction and weekly hours to numeric values. It calculated the total sample size, the arithmetic mean of satisfaction, the frequency and share of each 1–5 rating, and the count and share of each preferred feature. It then produced role-level sample sizes and satisfaction means, plus role-by-feature cross-tabulations with within-role shares. The tabular results were written to analysis.csv; charts and narrative findings were embedded in survey.html. The HTML uses only inline CSS and SVG and has no external dependency.
Actual results
Across the 12 records, the satisfaction total is 45 and the mean is 3.75. Rating 4 is the most frequent score, appearing five times; rating 2 appears twice, and rating 1 does not appear. Excel is the most preferred feature with 5 respondents (41.7%), followed by Documents with 4 (33.3%) and Code Review with 3 (25.0%). By role, Operations has 4 respondents and a mean satisfaction of 4.00, with Excel and Documents selected twice each. Development has 4 respondents and a mean of 3.50, with Code Review selected three times, Excel once, and Documents zero times. Teaching has 4 respondents and a mean of 3.75, with Excel and Documents selected twice each. Operations has the highest mean and Development the lowest, but every role contains only four records.
Checks
During generation, the script asserted that the source contains exactly 12 rows and recalculated every metric directly from the original CSV. Post-generation checks verify the sample size, mean, denominators, shares, feature labels, role cross-tabulation, and the small-sample warning in the output files. The HTML is self-contained: it contains no CDN links, external scripts, external fonts, or network requests, so it can be opened offline.
Limitations
The sample contains only 12 records and is synthetic, so it cannot represent the broader population. Each role has only four records, meaning that a single response can materially change a group result. No sampling frame, confidence intervals, significance tests, or open-ended responses are available. Therefore, these are descriptive results for the supplied demonstration data, not evidence of a real market-level preference or satisfaction difference. The feature question is treated as a single-choice field and does not establish a complete product prioritization.
Download and inspect the outputs
The archive contains the files created during this run. Use these checks when adapting the example to your own data.
- Reconcile the reported results against the downloadable sample inputs.
- Open the generated files and check the contents relevant to your task.
- Review the limitations and any unverified steps in the report.