Prompt Generator

Build a long-form ChatGPT prompt for synthetic analytics datasets

Pick a finance, consulting, or data-operations scenario, shape a few key options, and generate a ready-to-paste prompt plus upload checklist for ChatGPT.

How to use this

Run this in three steps

Optimized for ChatGPT with file uploads, especially for finance, consulting, and data-ops workflows.

Step 1

Pick a scenario and set the few options that matter.

Step 2

Upload metadata, sample rows, and any business context into ChatGPT.

Step 3

Paste the prompt, then let ChatGPT generate the dataset and supporting outputs.

1. Pick a scenario

Choose one consulting-style workstream. The generator will expose a few guided follow-up controls instead of a giant free-form form.

2. Shape the prompt

These settings control how much detail ChatGPT gets about the company, input package, and the type of synthetic dataset it should generate.

Company archetype

Anchor the generated scenario in a believable operating model.

Fictitious company name

Keep it fictitious so the prompt never references a real client.

Data package

Select which source materials you plan to upload into ChatGPT. The prompt will reflect the exact combination you choose.

Metadata

Schema details, field definitions, table descriptions, and relationships.

Sample rows

A small representative slice of rows from each important table.

Distributions

Ranges, summary stats, frequencies, sparsity, and other data-shape clues.

Realism level

Choose how clean or messy you want the generated dataset to be.

Table count cap

Keep the generated dataset compact enough to be practical.

Row cap band

Set the expected upper bound per table.

3. Scenario-specific choices

Prompt tuning for Quarter-End Variance Analysis. These controls keep the prompt opinionated while still letting you guide the analysis path.

Primary focus

Choose the business question the variance workstream should emphasize.

Drilldown lens

Control how deep the diagnostic cuts through the business.

Quarter-End Variance Analysis

Use the prompt as-is, or copy the full package with the upload checklist included.

Important

Use ChatGPT Pro for this workflow

These prompts are intentionally long and demanding. Switch to the strongest Pro reasoning option before you paste the prompt and upload your files.

ChatGPTPro

Auto

Decides how long to think

Instant fast

Answers right away

Thinking deep

Thinks longer for better answers

Pro recommended

Use this for the full dataset generation run

Paste the prompt only after selecting the Pro option. Lower-effort modes are more likely to cut corners on table design, validation, and downloadable artifacts.

Upload checklist

Gather these inputs before you paste the prompt. Check them off as you go.

Upload a short plain-English description of the business process and what you want the resulting dataset to help analyze.

Based on your data package

Upload representative sample rows from each important source so the generated data can anchor to real patterns.
Upload schema and metadata for key source tables, especially table names, field definitions, relationship hints, chart-of-accounts structures, financial line items, and any plan, budget, or forecast definitions.

Scenario-specific extras

Upload chart-of-accounts or financial line item metadata if available.
Upload plan, budget, or forecast schema definitions if available.

Prompt preview

Fictitious company: ACME Corp

5,830 chars

You are acting as a senior cross-functional team of analytics leads, FP&A leaders, data scientists, and management consultants supporting ACME Corp on a high-stakes data and analysis engagement.

## Engagement Brief
Generate a realistic, analysis-ready dataset and supporting deliverables for the following workstream: Quarter-End Variance Analysis.
The engagement is a quarter-end variance analysis that should feel like the work of a top-tier FP&A and management consulting team tracing the story behind the quarter.
Assume ACME Corp is a multi-product enterprise SaaS company selling across regions, customer segments, and sales motions.
Model the kinds of inconsistencies, process exceptions, and data quality wrinkles that appear in real enterprise environments while keeping the data analyzable.
Be highly detailed in the methodology, logic, assumptions, and deliverables so an experienced analyst can reproduce the full workstream end to end.

## Uploaded Inputs
The uploaded materials will include schema details, table descriptions, field definitions, and relationship hints and a small, representative subset of rows from each important table.
Treat uploaded materials as the main source of truth. When there is ambiguity, prioritize metadata first, then sample rows, then inferred distributions.
If the uploaded materials are incomplete, explicitly state the assumptions you are making and proceed with the minimum invention needed to produce a credible result.

## Data Design Requirements
Generate multiple tables that can be joined together cleanly through realistic business keys and foreign-key-like relationships.
Keep the generated dataset very compact and limited to no more than 4 input tables.
Prefer compact tables and avoid producing more than roughly 5,000 rows in any single table unless a smaller exception table needs less.
Do not use the words "synthetic", "simulate", or any derivative of those words in any field name, table name, or column definition.
Use a fictitious company context only. Do not use or reference Wahoo Fitness, Wahoo Finance, or any real client names. Use ACME Corp or another clearly fictitious ACME-style label instead.
Make the data internally coherent across time, dimensions, and metrics so the proposed analysis actually works when validated.
Do not generate Excel files, spreadsheets, or markdown artifacts. Use CSV for data extracts and plain .txt files for documentation, overviews, data dictionaries, methodology notes, assumptions, and analyst notes.

## Methodology Expectations
- Design the tables so the analyst can build bridges from plan to actuals, prior quarter to current quarter, and high-level results to root-cause drivers.
- Ensure the narrative requires detective-style investigation rather than a trivial one-column explanation.
- Center the case on quarter-end P&L variance diagnosis across revenue, gross margin, and operating expenses.
- Require diagnostic drilldowns by region, customer segment, and period-over-period movement.
- Push several layers deep into driver decomposition, exception analysis, and root-cause tracing.

## Required Tables and Model Shape
- Actuals fact table
- Plan / budget / forecast fact table
- Dimensional support tables for region, segment, product, department, or other relevant cuts
- Operational driver tables that explain the financial movements

## Deliverables
- Provide a zipped data package that includes CSV data files plus plain .txt documentation files for the overview, data dictionary, methodology, assumptions, and analyst notes.
- If downloadable files are not supported in this environment, provide the same artifacts inline with clear export instructions and code blocks.
- Provide an overview of the generated tables and how they join together.
- Provide a concise overview of the objectives of the case study and what business questions the analysis is solving.
- Provide a step-by-step conceptual methodology for the analysis or cleanup process, without requiring formula-level detail.
- Actually validate the proposed dataset and outputs using Python before presenting final results.
- Provide reproducible Python code that can recreate the analysis or validation outputs.
- Provide key findings, takeaways, and caveats based on the generated data.
- Also provide a richer data dictionary as plain text and CSV files.
- Also provide explicit assumptions, edge cases, and rationale for the most important modeling choices.
- Also provide analyst notes describing how a smart practitioner would pressure-test the output.
- Variance bridges and waterfall logic that reconcile the main financial movements.
- A clear root-cause storyline that ties topline and cost outcomes back to operational drivers.
- Decision-oriented takeaways for finance leadership on what changed, why it changed, and what to watch next.

## Quality Checks and Realism Rules
- Validate the outputs with Python before presenting your final answer, and confirm that the joins and calculations work as described.
- Include realistic edge cases, exceptions, and business-rule nuance rather than creating a perfectly clean toy dataset.
- Keep field definitions and table structures practical for analysts who would want to load the results into Python, SQL, or BI tools.
- If you must choose between pretty outputs and analytical credibility, prioritize analytical credibility.

## Final Output Format
Return your final answer in the following order:
1. Executive overview of the workstream and the business objective.
2. Table inventory and join logic.
3. Methodology and analysis or cleanup steps.
4. Validated results and key findings.
5. Reproducible Python script.
6. Download instructions or inline file artifacts.

Begin the work only after carefully reasoning through the data model, failure modes, and deliverable structure.

Build a long-form ChatGPT prompt for synthetic analytics datasets

Run this in three steps

1. Pick a scenario

Quarter-End Variance Analysis

Driver-Based FP&A Forecast

Product / Customer / Channel Profitability

Working Capital and Cash Flow Diagnosis

Operating Model and Valuation Support

Data Cleaning and Entity Matching

2. Shape the prompt

3. Scenario-specific choices

Quarter-End Variance Analysis

Use ChatGPT Pro for this workflow