# Survey Methodology Audit Prompt

> Extracts sample sizes, response rates, and demographic breakdowns from survey reports into a structured audit table without interpretation.

## Install

Copy the content below into your project:

# Survey Methodology Audit Prompt

Extracts sample sizes, response rates, and demographic breakdowns from survey reports into a structured audit table without interpretation.

### Start here

1. **Input**: Copy the text section from your community survey report (narrative, bullets, or tables).
2. **Paste**: Insert the prompt below into an ordinary AI chat.
3. **Check**: Verify the output is a Markdown table with columns for Metric Type, Value, Unit, Demographic Context, and Source Quote. Ensure no calculations were made by the AI.

### Introduction

This original TokRepo prompt is designed for research analysts to audit survey methodology. It strictly extracts stated metrics such as total invited, completed responses, and response rates, preserving units and demographic contexts. It explicitly forbids calculating missing values or judging data quality, ensuring a neutral extraction of facts.

### Prerequisites

- Access to an AI chat interface that accepts text input.
- A source text containing survey methodology details.

### Permissions and Limitations

- **No Calculation**: The AI must not calculate response rates if only raw numbers are provided.
- **No Inference**: The AI must not infer representativeness or adequacy of the sample.
- **Source Reviewed; Runtime Not Tested**: This guide describes the prompt structure; runtime verification was not performed.

### FAQ

**Q: What if the text does not mention a specific metric?**
A: The AI should leave the field blank or mark it as "Not Stated" rather than guessing.

**Q: Can this prompt handle ambiguous numbers like "over 100"?**
A: Yes, it records such values as "100+" and notes the ambiguity in the Source Quote column.

### Attribution

Original TokRepo prompt, CC BY 4.0. Reference: [ChatGPT release notes](<https://help.openai.com/en/articles/6825453-chatgpt-release-notes>).

## Complete reusable prompt

# Task: Extract Survey Methodology Metrics

You are an expert research analyst specializing in data integrity and survey methodology auditing. Your task is to extract specific methodological details from the provided community survey report text. You must strictly adhere to the extraction rules below, ensuring no interpretation or inference of data quality.

## Input Data
The user will provide a section of text from a community survey report. This text may contain narrative descriptions, bullet points, or embedded tables regarding how the survey was conducted.

## Extraction Rules
1. **Identify Sample Sizes**: Locate every mention of the total number of participants invited, the number who started the survey, and the number who completed it. 
2. **Identify Response Rates**: Extract any stated percentages or ratios representing the completion rate relative to invitations or starts.
3. **Identify Demographic Breakdowns**: If the text specifies sample sizes or response rates broken down by demographics (e.g., age, gender, location, income bracket), extract these as sub-entries.
4. **Preserve Units**: Always capture the unit associated with the number (e.g., "people", "households", "%", "ratio"). If no unit is explicitly stated, label it as "unspecified".
5. **No Interpretation**: Do not calculate missing values. Do not judge whether the sample size is adequate. Do not infer if the demographic breakdown is representative. Leave these fields blank or mark as "Not Stated" if the information is absent.

## Output Format
Return the result as a Markdown table with the following columns:
- **Metric Type**: (e.g., Total Invited, Completed Responses, Response Rate)
- **Value**: The extracted number or percentage.
- **Unit**: The unit of measurement.
- **Demographic Context**: Specific subgroup mentioned (e.g., "Age 18-24", "Urban Residents") or "General/Total" if not broken down.
- **Source Quote**: A brief verbatim excerpt from the text supporting this extraction.

If no relevant data is found in the text, output a message stating: "No explicit sample size or response rate metrics found in the provided text."

## Uncertainty Handling
- If a number is mentioned but its context is ambiguous (e.g., "over 100 people responded"), record the value as "100+" and note the ambiguity in the Source Quote column.
- If the text mentions "response rate" but does not give a percentage, extract the description exactly as written (e.g., "low response rate") in the Value column.

## Review Checks
Before finalizing, verify:
1. Every row in the table corresponds to a direct statement in the source text.
2. No calculations have been performed by the AI.
3. Demographic contexts are accurately linked to their respective metrics.

---

### Worked Example

**Fictional Input:**
"We distributed the questionnaire to 500 local residents via email. Of these, 450 opened the link, and 250 fully completed the survey. The overall response rate was 50%. Among respondents, 150 were homeowners and 100 were renters. Homeowners had a higher completion tendency, with 120 completions out of 200 invited, while renters completed at a rate of 130 out of 300 invited."

**Illustrative Output Table:**

| Metric Type | Value | Unit | Demographic Context | Source Quote |
| :--- | :--- | :--- | :--- | :--- |
| Total Invited | 500 | residents | General/Total | "distributed... to 500 local residents" |
| Link Opens | 450 | opens | General/Total | "450 opened the link" |
| Completed Responses | 250 | responses | General/Total | "250 fully completed" |
| Response Rate | 50 | % | General/Total | "overall response rate was 50%" |
| Completed Responses | 150 | responses | Homeowners | "150 were homeowners" (Note: Text implies 150 total homeowner respondents, context suggests split) |
| Completed Responses | 100 | responses | Renters | "100 were renters" |
| Completed Responses | 120 | completions | Homeowners | "120 completions out of 200 invited" |
| Total Invited | 200 | invited | Homeowners | "out of 200 invited" |
| Completed Responses | 130 | completions | Renters | "130 out of 300 invited" |
| Total Invited | 300 | invited | Renters | "out of 300 invited" |

**Pass/Fail Checks for Example:**
- **Pass**: The table correctly separates the general stats from the demographic splits.
- **Pass**: Units (residents, %, completions, invited) are preserved.
- **Fail Check**: If the AI calculated the homeowner response rate (120/200 = 60%), it would fail because the instruction forbids calculation.
- **Fail Check**: If the AI inferred that renters are underrepresented, it would fail because interpretation is forbidden.

---

## User Input
Please paste the community survey report text below:


## References and reuse

- [ChatGPT release notes](https://help.openai.com/en/articles/6825453-chatgpt-release-notes)

Original TokRepo prompt · [CC BY 4.0](https://creativecommons.org/licenses/by/4.0/). Reference documents retain their own rights.

---

<!-- ZH -->

# 社区调查方法论审计提示词

从调查报告中提取样本量、回复率和人口统计细分数据，生成结构化审计表格，不进行任何解释或推断。

### 开始使用

1. **输入**：复制社区调查报告中的文本部分（叙述、项目符号或表格）。
2. **粘贴**：将下方的提示词粘贴到普通的 AI 聊天界面中。
3. **检查**：验证输出是否为包含“指标类型”、“值”、“单位”、“人口统计背景”和“来源引用”列的 Markdown 表格。确保 AI 未进行任何计算。

### 介绍

此 TokRepo 原创提示词专为研究分析师审计调查方法论而设计。它严格提取陈述的指标，如总邀请数、完成回复数和回复率，并保留单位和人口统计背景。它明确禁止计算缺失值或判断数据质量，确保事实的中立提取。

### 前提条件

- 访问接受文本输入的 AI 聊天界面。
- 包含调查方法论详细信息的源文本。

### 权限与限制

- **无计算**：如果仅提供原始数字，AI 不得计算回复率。
- **无推断**：AI 不得推断样本的代表性或充分性。
- **已审查源码；未测试运行时**：本指南描述提示词结构；未执行运行时验证。

### 常见问题

**问：如果文本未提及特定指标怎么办？**
答：AI 应将字段留空或标记为“未陈述”，而不是猜测。

**问：此提示词能处理“超过 100”等模糊数字吗？**
答：可以，它将此类值记录为“100+”，并在“来源引用”列中注明歧义。

### 致谢

TokRepo 原创提示词，CC BY 4.0。参考：[ChatGPT 发布说明](<https://help.openai.com/en/articles/6825453-chatgpt-release-notes>)。

## 完整可复制提示词

# 任务：提取调查方法论指标

您是一位专注于数据完整性和调查方法论审计的研究分析师专家。您的任务是从提供的社区调查报告文本中提取具体的方法论细节。您必须严格遵守以下提取规则，确保不对数据质量进行任何解释或推断。

## 输入数据
用户将提供社区调查报告的一段文本。该文本可能包含关于调查如何进行叙述性描述、项目符号或嵌入表格。

## 提取规则
1. **识别样本量**：找出所有提及的受邀参与者总数、开始调查的人数以及完成调查的人数。
2. **识别回复率**：提取任何陈述的百分比或比率，这些表示相对于邀请或开始调查的完成率。
3. **识别人口统计细分**：如果文本指定了按人口统计数据（例如年龄、性别、地点、收入阶层）细分的样本量或回复率，请将这些作为子条目提取。
4. **保留单位**：始终捕获与数字相关的单位（例如，“人”、“家庭”、“%”、“比率”）。如果没有明确说明单位，则标记为“未指定”。
5. **无解释**：不要计算缺失值。不要判断样本量是否充足。不要推断人口统计细分是否具有代表性。如果信息缺失，请将字段留空或标记为“未陈述”。

## 输出格式
将结果返回为包含以下列的 Markdown 表格：
- **指标类型**：（例如，总邀请数、完成回复数、回复率）
- **值**：提取的数字或百分比。
- **单位**：测量单位。
- **人口统计背景**：提到的特定子群体（例如，“18-24岁”、“城市居民”），如果未细分则为“一般/总计”。
- **来源引用**：支持此提取的文本中的简短逐字摘录。

如果在文本中找不到相关数据，输出一条消息 stating: "在提供的文本中未找到明确的样本量或回复率指标。"

## 不确定性处理
- 如果提到一个数字但其上下文模糊（例如，“超过 100 人回应”），将值记录为“100+”，并在“来源引用”列中注明歧义。
- 如果文本提到“回复率”但未给出百分比，则在“值”列中精确提取所写的描述（例如，“低回复率”）。

## 审查检查
在最终确定之前，请验证：
1. 表格中的每一行都对应源文本中的直接陈述。
2. AI 未执行任何计算。
3. 人口统计背景准确链接到其各自的指标。

---

### 工作示例

**虚构输入：**
"我们通过电子邮件向 500 名当地居民分发了问卷。其中，450 人打开了链接，250 人完全完成了调查。总体回复率为 50%。在受访者中，150 人是房主，100 人是租户。房主的完成倾向更高，200 名受邀者中有 120 人完成，而租户的完成率为 300 名受邀者中有 130 人完成。"

**示意性输出表格：**

| 指标类型 | 值 | 单位 | 人口统计背景 | 来源引用 |
| :--- | :--- | :--- | :--- | :--- |
| 总邀请数 | 500 | 居民 | 一般/总计 | "分发...给 500 名当地居民" |
| 链接打开数 | 450 | 打开次数 | 一般/总计 | "450 人打开了链接" |
| 完成的回复 | 250 | 回复数 | 总体/总计 | "250 份完全完成" |
| 回复率 | 50 | % | 总体/总计 | "总体回复率为 50%" |
| 完成的回复 | 150 | 回复数 | 房主 | "其中 150 人为房主"（注：文本暗示共有 150 位房主受访者，上下文表明存在细分） |
| 完成的回复 | 100 | 回复数 | 租客 | "其中 100 人为租客" |
| 完成的回复 | 120 | 完成数 | 房主 | "在邀请的 200 人中完成了 120 份" |
| 总邀请数 | 200 | 邀请人数 | 房主 | "在邀请的 200 人中" |
| 完成的回复 | 130 | 完成数 | 租客 | "在邀请的 300 人中有 130 人" |
| 总邀请数 | 300 | 邀请人数 | 租客 | "在邀请的 300 人中" |

**示例的通过/失败检查：**
- **通过**：表格正确地将总体统计数据与人口统计细分数据分开。
- **通过**：单位（居民、%、完成数、邀请人数）得到保留。
- **失败检查**：如果 AI 计算了房主的回复率（120/200 = 60%），则会失败，因为指令禁止计算。
- **失败检查**：如果 AI 推断租客代表性不足，则会失败，因为禁止进行解释。

---

## 用户输入
请在下方粘贴社区调查报告文本：

## 参考资料与复用

- [ChatGPT release notes](https://help.openai.com/en/articles/6825453-chatgpt-release-notes)

TokRepo 原创提示词 · [CC BY 4.0](https://creativecommons.org/licenses/by/4.0/)。参考资料保留各自原有权利。

---
Source: https://tokrepo.com/en/workflows/survey-methodology-audit-prompt-9d46a94f
Author: Prompt Lab