# Fictional English-rule change evaluation

Status: Not tested. Expected decisions below are hand-calculated policy expectations, not Kognitos execution results. These records are separate from the on-page SOP/journal example and from the Finance six-case source pack. No connected run, customer outcome, savings or timing is asserted.

## Scope and setup

Use the CSV as synthetic scalar input, not an ERP import or extracted invoice. PO net value is USD 1,000.00. Invoice values exclude tax and freight, use the same currency and have cent precision. Assume one PO and invoice per case; no aggregation or partial receipts. Only the net amount comparison is in scope. Vendor, duplicate, receipt, tax, access and posting checks are outside this exercise; 'Within tolerance' never means approved to pay.

Use an isolated test environment with no ERP posting/payment action. A technical evaluator supplies the required input mapping and confirms which actual procedure/version is run. If a setup or evidence field is unavailable, mark the case Blocked and explain why; never enter the expected value as the actual result.

## Policy before the edit — 1%

If the purchase-order total or invoice total is missing, hold the invoice for review. Require a positive purchase-order net total. Otherwise, hold the invoice when the absolute difference between the invoice net total and the purchase-order net total is greater than 1% of the purchase-order net total. At exactly 1%, record Within tolerance. Do not post or pay.

## Proposed policy after the edit — 2%

Change only the percentage to 2%. Keep missing-data behavior, comparison base, currency, strict greater-than operator and no-write boundary unchanged. At exactly 2%, record Within tolerance. These are readable evaluation policies; validate the actual automation document in the environment rather than treating this worksheet as executable syntax.

## Expected decisions to review

| Case | Invoice net USD | Before 1% | After 2% |
|---|---:|---|---|
| RC01 | 1,005.00 | Within tolerance | Within tolerance |
| RC02 | 1,010.00 | Within tolerance | Within tolerance |
| RC03 | 1,015.00 | Hold for review | Within tolerance |
| RC04 | 1,020.00 | Hold for review | Within tolerance |
| RC05 | 1,020.10 | Hold for review | Hold for review |
| RC06 | Missing | Hold for review | Hold for review |

Calculate using exact decimal amounts: abs(invoice - PO) > PO × threshold. Do not round the percentage before comparison. All six cases must be repeated in both versions. The changed RC03/RC04 decisions and unchanged boundary/missing cases are the regression check. Add negative variances and invalid inputs before any wider deployment; these six cases alone are not a full validation suite.

## Evidence to fill for each case and version

- Case ID:
- Test environment and date:
- Business policy owner and approval for the proposed change:
- Actual source/input record and checksum:
- Actual procedure/version ID and approved English document:
- Run ID and timestamp:
- Actual decision (leave blank until observed):
- Link to run evidence and step used:
- Expected versus actual comparison:
- Confirmation no ERP write/payment occurred:
- Technical assistance and elapsed edit/test time:
- Reviewer, discrepancies and resolution:
- Status: Not tested / Blocked / Passed / Failed

## Release decision

Resolve failures, review connected inputs and permitted actions, and obtain the required approval before publishing a tested version. Preserve the baseline run and version evidence. Record the released version separately; changing the worksheet does not change a published automation.

Documentation: https://docs.kognitos.com/guides/platform/drafts (retrieved October 26, 2026). Kognitos documents editing a published automation in Drafts, testing, and publishing a new major version. Actual run and deployment behavior remain to be verified in the selected environment.