Mask Financial Data & Ratios for AI
Proportional ×kMask financial data for AI—protecting revenue, margins, IBANs, and tax IDs via proportional ×k scaling and date shifts so LLMs compute true ratios.
- Proportional ×k preserves ratios
- IBAN / Luhn / Tax ID validated
- Reversible with rounding check
Instant Before & After Preview
Proportional ×k preserves growth ratiosQ1 Treasury Brief (2026-01-15 to 2026-04-15)
Revenue: $1,200,000 · Gross Profit: $360,000 (30.0% margin)
Settlement IBAN: DE89 3704 0044 0532 0130 00 · VAT: DE123456789Q1 Treasury Brief (2026-03-25 to 2026-06-23)
Revenue: $1,944,000 · Gross Profit: $583,200 (30.0% margin)
Settlement IBAN: [[IBAN-01]] · VAT: [[TAX-01]]How it works
How proportional scaling hides real figures without breaking AI math
Replacing dollar amounts with [[AMT-01]] stops AI from checking gross margins or YoY growth. Scaling every figure by one secret factor solves both.
- Paste financial report or tableCurrency figures ($1,250,000, €120,000, ¥500,000, 100万元), IBANs, bank cards, and dates are detected locally.
- Scale by secret factor ×kEvery amount multiplies by a non-integer session factor (e.g. ×0.73) and dates shift uniformly so gross margin and growth % stay identical.
- Reverse AI figures with rounding reviewPaste the AI analysis back into Restore—exact mapped amounts revert directly and derived figures show amber rounding review hints.
FAQ
Frequently asked questions
Why does proportional scaling keep gross margin and YoY growth percentages intact?
Multiplying revenue and COGS by the same constant k leaves (Revenue − COGS) / Revenue and year-over-year percentage ratios mathematically unchanged.
What happens when restoring numbers that the AI computed from scaled figures?
Known scaled figures revert to exact originals, while newly derived figures reverse via ÷k and highlight in amber so you can verify rounding.
Can I re-roll the random date shift or scale multiplier?
Yes. Click Re-roll in the fidelity controls bar to generate a fresh non-zero day offset or scale multiplier anytime before sending to AI.