# Translate a recap without changing the claim

Original constructed practice and complete working route. No real participant response, product test, actual incident or customer evidence. Any prior first-party evidence in the existing guide retains only its expressly stated scope.

Translate meaning and presentation separately. Preserve the measured unit, denominator, period, first-person perspective and uncertainty, then localize numbers and unambiguous dates. Fluent wording cannot turn impressions into readers, response events into people, or collected posts into a complete publishing history.

## Write a small claim record before translating

Separate what was observed from the sentence you want to publish. Record the actual quantity, unit, denominator, time window and what remains unknown. Keep a personal “I recorded” statement personal rather than changing it into an organization’s achievement or a universal recommendation.

Use the original text privately and avoid sending sensitive account details into an unnecessary public example. The downloadable cases are constructed and contain no customer or reader observations. Their purpose is to make semantic changes visible.

## Review the complete 30/1,200 example

The safe English fixture says: “In this constructed October 2026 UTC example, I recorded 30 response events across 1,200 impressions (2.5%). These are events, not different people.” The complete German version preserves every claim and uses 1.200 and 2,5 %.

The calculation is 30 ÷ 1,200 × 100 = 2.5%. The comma and point change with presentation, not with quantity. Neither version calls impressions readers or claims 30 different respondents. Keep the source definition of response events; do not substitute an appealing marketing label.

## Catch a fluent but unsupported translation

The packet deliberately labels an unsafe version: “I reached 1,200 readers and converted 2.5% into customers.” It sounds plausible but changes display events into people and a response-event ratio into a customer result. No translation polish can repair those missing observations.

Compare the factual nouns and verbs, not only the percentage. Ask whether reached, read, responded, converted, collected and published retain their original meaning. Remove or rephrase an unsupported claim in both versions instead of letting a confident second-language sentence become the new authority.

## Keep collection differences out of growth claims

A second constructed example has 12 public posts in one collection and 18 in another for the same period. A safe bilingual statement says the second collection found six more posts. It does not say publishing rose by 50%, that all posts were recovered, or that the creator became more active.

Translate the collection qualifier together with the numbers. If space is tight, cut a decorative introduction before cutting the limitation. A reader should not need the other language or an unavailable image to discover the actual scope of the claim.

## Make the time window readable in both languages

Use October 2026 / Oktober 2026 with an explicit UTC window from October 1 inclusive to November 1 exclusive for this exercise. Avoid an ambiguous numeric date whose month and day could be reversed. Preserve the actual timezone instead of adapting it silently to the translator’s location.

W3C’s internationalization guidance addresses language and local formats, while its date guidance explains ambiguity. Those presentation principles do not certify a translation’s factual accuracy. Check the complete sentence after number formatting, because units or qualifiers can still have disappeared.

## Compare the complete edition before handoff

For a bilingual handoff, keep one shared meaning record and separate presentation choices. In constructed edition L2, numeric value 1250.5 stays unchanged; the reading versions may show 1,250.5 and 1.250,5. Date 2026-11-04 may appear as 4 November 2026 and 4. November 2026. Neither the day nor the actual UTC boundary changes. A fluent sentence does not replace this factual comparison.

Three additional constructed checks: L1 changes “not measured as customers” into a customer total; restore the negation and unknown effect in both versions. L2 confuses 4 November with 11 April; inspect the unambiguous source instead of guessing an ordering. L3 translates a machine-readable field key even though the importer expects its original identity; keep keys and data separate from translated explanations. None of these checks is an actual participant response.

W3C separates localized display from preferably locale-neutral data representations; localized values need suitable context. Our handoff record keeps source edition, both full texts, quantities/units, time basis, negations, actual link destinations and remaining uncertainty together. Inspect alternatives and notes outside the main prose too. The new worksheet replaces neither professional language review nor the older original files.

| Check object | Preserve | Hold or correct |
| --- | --- | --- |
| Unit | Response events | Do not rename as people or customers |
| Quantity | 1250.5 as shared numeric value | Check localized display separately |
| Date | 2026-11-04 and actual UTC basis | Do not guess a day/month order |
| Negation | Not measured remains not measured | Do not invent a positive effect |
| Field key | impressions as declared identity | Localize explanation rather than the key |
| Edition/destination | Shared L2 edition; actual locale links | Do not pass unverified destinations |

All L-cases and this handoff are constructed; no real translation test or new reader.

## Run arithmetic, then do the meaning review

The local checker confirms 30/1,200 and the English/German percentage formatting, plus original file identities. Run node reproduce-examples.mjs from the unzipped packet with Node.js 22 or newer already installed. It makes no account action and does not score language quality.

Complete the translation record with original meaning, quantities, dates, limits and the unresolved bilingual question. Read both full versions as standalone posts, including image alternatives and next steps where present. Finish only when neither requires the other language to correct a stronger claim.

## Sources and limits

- [W3C: unambiguous international dates](https://www.w3.org/International/questions/qa-date-format)
- [W3C: internationalization, language and localized formats](https://www.w3.org/International/quicktips/)
- [Original complete worked example (JSON)](https://wrapped.li/blog-assets/editorial-workflows/translation-cases.json)
- [Original constructed examples, bilingual review records and local checker (ZIP)](https://wrapped.li/blog-assets/editorial-workflows/complete-editorial-workflows.zip)
- [W3C: metadata, versioning and locale-neutral representations (declared scope)](https://www.w3.org/TR/dwbp/)