Clean Data. Better Decisions.
Sales Data Quality Auditor
Profile, dedupe, validate, export
Download FreeFeatures List
8
Structured Prompts
19
Data Fields Covered
100%
Free to Download
Why Finance and RevOps Teams Use This Prompt Book
8 Prompts, One Workflow
Each prompt covers a distinct stage of the audit: profiling, standardisation, deduplication, anomaly detection, missing data handling, validation, change logging, and final export. Run them in sequence for a complete, traceable audit trail.
Catches What Spreadsheets Miss
The prompts apply finance-grade logic: IQR-based outlier detection per product, ISO 4217 currency normalisation, fuzzy account name matching at 0.85 threshold, and a math check on total_amount vs. extended_amount minus discount plus tax.
Audit-Ready Change Log
Prompt 7 generates a structured change log with change_id, rule version, approver, decision, timestamp, and a hash of the input file. Every data change is documented for audit defensibility — not just noted in a comment.
Copy, Paste, Done
No setup. No login. No configuration. Open ChatGPT, upload your sales file, and paste the first prompt. The book tells you what to expect from each prompt, what decision to make, and what to do next.
CFO-Ready Final Report
Prompt 8 produces a before/after scorecard with duplicate rate, missingness by column, standardisation coverage, validation pass rate, and a three-bullet executive narrative written for the CFO.
What's Inside the Free Download
-
Prompt Book PDF (8 Prompts)
A structured, copy-paste-ready prompt sequence for running a full sales data audit in ChatGPT. Covers profiling, standardisation, deduplication, outlier detection, missing data handling, validation, audit logging, and final QA export.
What Each Prompt Does — and What You Get Back
Prompt: Profiling
What You Ask ChatGPT: Upload your file. Ask ChatGPT to infer data types, report nulls, list suspect values, and propose a 19-field schema mapping.
Output You Receive: Your company has grown, but the expense policy hasn't changed since the early days. Or there is no policy — just informal rules and manager judgment. The Architect assesses what you have, flags the gaps, and produces a policy that fits where the business is now.
Prompt: Standardisation
What You Ask ChatGPT: Ask ChatGPT to propose rules for dates (ISO 8601), currency codes (ISO 4217), country names, account name casing, and SKU format.
Output You Receive: Rules table with logic, before/after examples, and confidence rating (H/M/L). You approve the rules.
Prompt: Deduplication
What You Ask ChatGPT: Define exact-match, near-match, and survivorship rules. Fuzzy account name match at 0.85 threshold, 24-hour temporal window.
Output You Receive: Rules summary, SQL-like pseudocode, expected duplicate counts by rule.
Prompt: Outlier Detection
What You Ask ChatGPT: Run IQR-based outlier detection per product, 3-sigma checks on extended_amount and margins, date range checks, geo-region alignment.
Output You Receive: Rule-by-rule output: affected rows, 5 sample rows, and a recommended action (flag / correct / ignore).
Prompt: Missing Data
What You Ask ChatGPT: Set policies by field type: flag-only for financial amounts, interpolation for bounded dates, mode-fill for categoricals at threshold >= 0.6, forward-fill for account attributes.
Output You Receive: Policy table: field, handling rule, pseudocode, rows affected, risk note.
Prompt: Validation
What You Ask ChatGPT: Run four business-rule tests: amount math check (tolerance 0.01), period completeness, credit limit breach, and margin check.
Output You Receive: PASS/FAIL results with counts, 5 sample fail rows, and a proposed fix or exception for each rule.
Prompt: Audit Logging
What You Ask ChatGPT: Consolidate all changes into a structured change log with rule_id, version, approver field, decision, ISO 8601 timestamp, and file hash.
Output You Receive: CSV-formatted change log rows plus a human-readable summary of key decisions.
Prompt: Final QA & Export
What You Ask ChatGPT: Produce a before/after scorecard across duplicate rate, missingness, standardisation coverage, validation pass rate, and anomaly rows.
Output You Receive: Markdown scorecard tables, 3-bullet CFO narrative, BI/ERP export checklist, and a Next Run checklist.
What Changes When You Work from a Structured Prompt Sequence
| Dimension | Ad Hoc ChatGPT Cleaning | Sales Data Quality Auditor |
|---|---|---|
| Coverage | Varies by what you think to ask | 8 stages, all dimensions covered |
| Deduplication logic | Simple exact-match at best | Exact + fuzzy match (0.85) + survivorship rules |
| Audit trail | None — chat history only | Structured change log with rule ID, timestamp, file hash |
| Outlier detection | Manual eyeballing | IQR per product + 3-sigma on amounts and margins |
| Final report | No standard output | Before/after scorecard + CFO narrative + export checklist |
| Cost | Free — but inconsistent every time | Free — same rigorous process every time |
Common Questions
Do I need a paid ChatGPT plan?
What format does my sales data need to be in?
Do I need to run all 8 prompts every time?
Is my data safe when I upload it to ChatGPT?
What's the difference between this and a paid AI for CFO product?
Can I use this for data other than sales?
Your Sales Data Has Problems. These 8 Prompts Find All of Them.
Updated for 2026. A full audit workflow for ChatGPT — free, structured, and audit-ready.