# Stop Excel from changing the identifiers your agent imports

Native Excel changed leading-zero codes and 16-digit document IDs in our CSV test. Explicit text preserved them. Includes the two-row fixture, files and tested import script.

Published 2026-11-15T07:20:00Z by Renaissance Superintelligence Corporation (RSI).

Choose identifier types before writing cells. In our native Excel CSV test, automatic opening turned entity 00100 into 100 and document 1234567890123456 into 1234567890123450. Writing identifier columns as text first preserved every character while keeping the amount numeric.

## An identifier can look numeric without being a number

Our fixture has Entity, Account, Document and Amount columns. The first record uses entity 00100, account 010100, document 1234567890123456 and amount 1250.25. The second uses 00200, 020100, 9999999999999999 and 2400.50. These are fictional codes and document identifiers, not personal or payment information.

The amounts are quantities we intend to add. The other columns identify records. Their characters matter: changing a code or a document key can break a join, prevent a match or merge records that should remain distinct. A financial total can still tie while those identities are wrong.

Microsoft documents automatic conversions and Excel’s 15-digit numerical precision. It recommends text for identifiers that need their original characters preserved. Our experiment measures the behavior in the tested desktop configuration; current Excel also offers conversion settings that can change the automatic-opening path.

## The amounts survived while the keys changed

We opened the CSV through Workbooks.Open and read Value2 from the cells. Entity and account codes lost their leading zeros. Both 16-digit document identifiers changed. The amounts remained 1250.25 and 2400.5, so an amount-only acceptance check would have missed the identity damage.

In the second path, we parsed the CSV as strings, formatted the first three destination columns as text, and wrote those fields as explicit strings. Amount was cast separately to a double. The codes and all 16 document characters then matched the source.

First fixture record: exact native cell values

Field | Source | Automatic open | Explicit types

Entity | 00100 | 100 | 00100

Account | 010100 | 10100 | 010100

Document | 1234567890123456 | 1234567890123450 | 1234567890123456

Amount | 1250.25 | 1250.25 | 1250.25

## Set the type before the value arrives

The critical ordering in the tested path is to apply text formatting to identifier columns before setting their values. Casting the imported PowerShell fields to strings also avoids handing wrapped CSV values directly to COM. Monetary cells receive a separate numeric conversion.

```powershell
$sheet.Range('A:C').NumberFormat = '@'
$sheet.Cells.Item($row, 1).Value2 = [string]$record.Entity
$sheet.Cells.Item($row, 2).Value2 = [string]$record.Account
$sheet.Cells.Item($row, 3).Value2 = [string]$record.Document
$sheet.Cells.Item($row, 4).Value2 = [double]$record.Amount
```

![Actual Excel typed import showing leading-zero entity and account codes and complete sixteen-digit document identifiers](/research/native-office-2026-11-15/identifiers.png)

The saved typed workbook reopened read-only on Panes. Cropped native desktop screenshot; the source values are also available as accessible text above.

## Compare keys as well as totals

Treat the raw export as the authority for the identifier characters. Compare the imported key set with the source key set, inspect duplicates and confirm that expected joins resolve. Do not normalize leading zeros away unless the source’s identity contract explicitly allows it.

A custom number format can make a short numeric code display leading zeros. That is different from preserving a text identifier for downstream programs. Formatting an already-truncated long number does not recover the missing digit. Reimport from the intact source instead of guessing a repair.

Our numeric amount conversion matches this fixture’s simple decimal-point CSV. Production imports need an explicit locale, currency and precision policy. The duplicate guard in our companion guide uses Decimal for source comparisons; its normalization rules should likewise be checked against the source contract.

## Run both paths against the same bytes

Download the [two-row CSV](/research/native-office-2026-11-15/identifier-fixture.csv), [automatic workbook](/research/native-office-2026-11-15/identifiers-automatic.xlsx), [typed workbook](/research/native-office-2026-11-15/identifiers-typed.xlsx), [result JSON](/research/native-office-2026-11-15/csv-identifiers.json) and [native script](/research/native-office-2026-11-15/native-csv-identifiers.ps1). Run the script in a new task directory. It creates its own fictional CSV and saves separate outputs.

The test is deliberately small so every changed character is inspectable. It establishes this failure mode, not a throughput result or comprehensive CSV parser evaluation. Add your own delimiter, quoting, encoding and locale cases before accepting a production source.

For an agent working in [real Excel on Panes](/use-cases/excel-automation), identity checks belong before query refresh and report delivery. Pair them with the [duplicate-document guard](/blog/power-query-duplicate-document-guard) and [independent workbook validation](/blog/verify-ai-generated-excel-workbooks).

## Evidence

Prepared native execution on one retained 4-vCPU, 16-GB Windows research computer, Excel 16.0 build 20527, November 15, 2026. Two fictional records, two import paths. The script reached its saved outputs in 1.634 seconds before application cleanup; API time was 2.413 seconds and controller wall time 2.499 seconds. These are observations of this prepared test, not an autonomous benchmark. No per-task operating cost is measured. Public workbook author metadata was redacted separately and both copies reopened read-only in native Excel without changing their bytes.

## Sources

- [Microsoft: leading zeros and large numbers](https://support.microsoft.com/en-us/excel/keeping-leading-zeros-and-large-numbers)

- [Exact native import results](https://usepanes.com/research/native-office-2026-11-15/csv-identifiers.json)

- [Public native reopen evidence](https://usepanes.com/research/native-office-2026-11-15/publication-reopen.json)