Translations can, and do, go wrong.

And when they do, the consequences are not just linguistic.
They can fundamentally distort your data.

The problem: when data and reality don’t align

A mobile health clinic operating in a small town on market days was evaluated using a standard 5-point Likert scale.

Respondents were asked to rate key aspects of service delivery:

• Waiting time
• Privacy
• Comfort in asking questions
• Customer service
• Perceived quality
• Reliability

The results raised concern.

Overall scores were low, and reliability stood out as particularly weak.

On paper, this suggested a serious performance issue.

But something didn’t add up.

The contradiction: positive experiences, negative scores

A follow-up qualitative study was conducted to explore the findings further.

Participants were asked to map local health facilities against their “ideal” clinic—visually indicating how each one compared.

The results told a very different story.

The mobile clinic—referred to as Clinic P—was consistently rated above average.

Participants spoke positively about:

• Their experiences
• The quality of care
• Their interactions with staff

So why had the quantitative data told the opposite story?

What went wrong: a failure of meaning, not measurement

The issue was not in the methodology.

It was in the translation.

The term “reliable”, in English, is commonly understood as:

consistent in quality, dependable, trustworthy.

But in the local language, its closest equivalent carried a very different meaning:

something that is always there—like a natural spring that never runs dry.

For a mobile clinic that only operated on specific market days, this interpretation didn’t fit.

Respondents weren’t questioning the quality of care.

They were responding to the clinic’s physical absence outside its operating schedule.

The result: a misleading dataset

What appeared to be poor performance was, in reality:

• A mismatch in meaning
• A contextual misunderstanding
• A translation that shifted the intent of the question

The dataset didn’t reflect dissatisfaction.

It reflected semantic misalignment.

What this reveals

In multilingual and multicultural contexts, translation is not just technical; it is interpretive.

Many local languages:

• Do not have direct equivalents for abstract descriptors
• Attach meaning through context, not definition
• Anchor concepts in lived experience rather than standardized terminology

This creates a critical risk:

You may think you are measuring one thing while respondents are answering something entirely different.

What this means for organizations

For organizations relying on research to guide decisions, this has serious implications:

• Flawed data can lead to incorrect conclusions
• Strong programs may be misjudged as underperforming
• Strategic decisions may be based on misunderstood signals

Translation is not a final step in the process.

It is a core part of research design.

At GMaurich, we don’t just translate words; we interpret meaning within context.

Because in research, accuracy is not just about what is asked—

It’s about what is understood.

Closing thought

Data is only as reliable as the meaning behind it.

And sometimes, the biggest risks are not in what people say—

but in what we think they mean.