Most safety reporting tools now list the languages they support. From the list you learn what the software accepts. You don’t learn whether a worker can explain what happened in that language, understand the follow-up questions and recognise their own account in the finished record.
In The Frontline Reporting Index 2027 we wrote that “A language option alone is not evidence that all three happen reliably.” This article is about how to test it, on your sites, before you rely on it.
Why isn’t a language option enough?
Because a report in a second language has to work in both directions. The worker has to get the account in, follow the questions that come back, and check the result before it carries their name. A tool can accept the language at the first step and still fail at the second or the third.
We opened the language chapter of the report with the point we’d make to any operations leader: “Language is an operational issue, not just a feature.”
Are language and literacy the same problem?
No, and treating them as one leads to the wrong fix. A worker can speak a language fluently and still find it hard to write, in that language or any other. Translating a fifteen-field form gives that worker a translated fifteen-field form.
That is one reason the choice of voice or text matters. Some workers will want to speak, some to type, and the same worker may switch depending on the noise and who is standing nearby. We went through the language side of training and reporting in safety training in a language workers understand.
What does the Assistance-Required Rate measure?
How often a worker needs help to report. In the report it is one of eight measures: “Reporting attempts needing help due to language, access, complexity or uncertainty. Percent of attempts.”
It is more useful read next to a second measure, the Observer-Initiated Reporting Rate, the share of accepted records started by the person who saw the event. With both, you can see whether the people with the information can get it into the system themselves, and where someone else steps in. The report’s guidance for a pilot is to “Preserve the distinction between who initiated a report, who helped and who confirmed it.”
In practice, that means recording three things for each attempt: who started it, whether anyone helped and why, and who confirmed the final account. In the report we also suggest you “Record when help is required, when an attempt is abandoned and when the worker corrects the account.”
How do you test meaning in another language?
With bilingual reviewers, in the conditions where the reporting will happen.
The report’s guidance: “Check meaning with competent bilingual review for the languages being evaluated. Test the actual conditions of use, including noise, device access and relevant vocabulary.”
What that looks like in a pilot:
- Choose the languages your crews speak, not the ones on the vendor’s list.
- Use a bilingual reviewer who knows the work. They compare what the worker said with what ended up in the record, and flag anything lost or changed.
- Test where the work happens. Noise, PPE, poor signal and shared devices all affect whether the exchange works.
- Test your vocabulary. Equipment names, site slang and the words crews mix between languages.
- Count the failures. Abandoned attempts and corrections are the evidence, so keep them in the sample.
One more line from the report belongs in the test plan: “Translation accuracy requires a separate check.” Check the accuracy of the translation separately from whether the worker could take part.
What do regulators require on language?
Mostly, they cover what the employer says to the worker. In its training policy statement, OSHA writes that training “must be presented in a manner that employees can understand”, and that “an employer must instruct its employees using both a language and vocabulary that the employees can understand.”
There is no general equivalent for the report travelling the other way, from the worker to the employer. Whatever language the reporting path accepts is usually decided by how the system was configured.
What should you ask a vendor, including us?
Six questions that separate a language list from a tested capability.
- Which languages have you tested in conditions like ours: noise, PPE, our equipment names?
- Who checked meaning, and were they competent in both languages and in the work?
- How does the worker check the finished account, in their language, before it is recorded?
- What happens when the system isn’t sure what was said?
- Does the record include who started it, who helped and who confirmed it?
- What do you do when a site needs a language you haven’t tested?
Where does Tandm fit?
Tandm talks with the worker in their own language, by voice or text. It asks the follow-up questions in that language and reads the account back for the worker to confirm before anything is recorded in their name. The confirmed record goes into the safety system you already run.
In a pilot, we’d expect you to test meaning as we set out in the report, with your own bilingual reviewers and your own vocabulary.
Book a demo and speak a report yourself, in whichever language you like.
Sources
- Tandm, The Frontline Reporting Index 2027: The work-to-record gap, framework edition, October 2026, chapter 4 and page 16. tandm.io
- Occupational Safety and Health Administration, Training Standards Policy Statement, 28 April 2010. osha.gov/training/compliance