Skip to content
Sign inRun the verification
How it works

One real workbook, from formulas to a verified system.

Every step below happened when this page was built, on the same example workbook you can open at /try. The numbers are the product’s output, not ours.

Run the live verification Talk to us about a pilot
Step 1 · Read

Understand what the workbook is for.

Tavuni reads the file’s own parts — formulas, the values Excel stored for them, named ranges, tables, hidden sheets — and works out what each sheet is: a ledger of records, a lookup table, a settings tab, a report. It is given no knowledge of this workbook.

10 sheets (1 hidden), 2,993 formulas, 2,993 of them supported by the formula engine.

The 'What we found in your workbook' screen: sheets, formulas, formulas we can run and logic we can rebuild, with a breakdown of supported formulas.
What we found — each sheet classified, with its confidence and reasoning.
Step 2 · Model

Build the Business Graph.

7 record types, 6 relationships — each relationship found from the lookup formulas that already join the sheets, with its evidence and confidence kept.

TaxRate6 records · 3 fieldsPricingRuleMinSubt…5 records · 3 fieldsPricingRuleUplift4 records · 2 fieldsCustomer40 records · 8 fieldsProduct35 records · 9 fieldsOrder120 records · 14 fieldsOrderLine380 records · 9 fields
Arrows point from the record that holds a reference to the record it refers to. Found from the workbook’s own lookup formulas — not drawn by hand.
The 6 relationships as a list
  • Customer.segment → PricingRuleUplift.segment
  • Customer.country → TaxRate.country
  • Order.customerId → Customer.customerId
  • OrderLine.orderId → Order.orderId
  • Order.shipCountry → TaxRate.country
  • OrderLine.sku → Product.sku
Step 3 · Compile

Turn formulas into rules about the business.

18 calculated fields, each compiled from thousands of copied-down cells into one rule that names entities and fields. The original formula stays beside it, so any number traces back to the cell it came from.

More on the model
Order.subtotal
Excel
=SUMIFS(OrderLines!$H$4:$H$383,OrderLines!$B$4:$B$383,A4)
Rule
SUM(OrderLine.lineTotal WHERE OrderLine.orderId = Order.orderId)
Order.taxRate
Excel
=IFERROR(VLOOKUP(E4,TaxRates!$A$4:$B$9,2,FALSE),DefaultTaxRate)
Rule
IFERROR(LOOKUP(TaxRate WHERE TaxRate.country = Order.shipCountry).taxRate, Setting.DefaultTaxRate)
Order.freight
Excel
=IF(D4="Cancelled",0,MAX(FreightMinimum,ROUND(I4*FreightRate,2)))
Rule
IF(Order.status = "Cancelled", 0, MAX(Setting.FreightMinimum, ROUND(Order.netAmount * Setting.FreightRate, 2)))
Step 4 · Verify

Check two ways — and show every difference.

The formula engine: 2,993 formulas recalculated and compared with the values the workbook stored — 2,992 exact, 1 on a rounding boundary, 0 different.

The generated system: the Business Graph run over 590 extracted records with calculated fields blanked, so the engine must derive every one. No cell addresses are involved — a wrong entity boundary, field mapping or relationship would show up here.

The expected values in this example workbook were computed by an independent, third-party Python formula engine — a different language and a different author — so the check is not the product grading itself.

Does the new system agree with the spreadsheet?northwind-ops.xlsx
Values checked
2,870
Matched
2,869
Rounding edge
1
Different
0

The generated system, run on 590 records with every calculated field blanked first, agrees with the workbook’s own stored results on 100.00% of values. The one value sitting exactly on a rounding midpoint is reported separately, never folded into the pass rate.

Step 5 · Evolve

Change it, safely.

A plain-language request becomes operations on the model, is validated, applied to a copy, run against the real records and shown as a diff. Only a person can publish it as the next version.

In this public example the proposal comes from deterministic code (local-heuristic (rule-based-v1)); no language model is called.

Safe change in detail
  1. 1 · Request
    Add a manager approval when order value exceeds $10,000.
  2. 2 · Proposal
    Require Manager approval for orders over $10,000Assumptions it states, so a person can disagree:
    • The threshold applies to Order.orderTotal (Order Total).
    • Reusing the existing "Manager" role.
    • The comparison is strictly greater than the threshold.
  3. 3 · Diff
    • addedAdded rule "Orders over $10,000 need Manager approval": Order.orderTotal > 10000money or approvals
    • addedAdded process "Order approval" with 4 stepsmoney or approvals
  4. 4 · Effect on real records
    89 of 120 Orders records (74.2%) would be affected.Combined order total of the affected records: $3,566,077.
  5. 5 · A person decides
    The assistant cannot publish. A person approves — and the change becomes version 2, with its author recorded — or discards it.
After the compiler

The same model runs the platform.

Beyond a workbook, the Business Graph is configured directly — business objects, stages, workflows, approvals and roles — and the platform generates the application people work in.

Tour the platform

Run this exact example yourself.

The example operations workbook opens in Tavuni with no account and nothing uploaded. Every number on this site comes from that same pipeline.

Run the live verification Talk to us about a pilot