By Ritvars Mētra, Founder of ReadinessCompassThe readiness survey promises anonymity. It also asks for function, location, role level and team, because the programme needs to know where the risk sits.
In a team of six, those four fields identify everyone. The respondents work this out considerably faster than the programme does, and they answer accordingly.
Anonymous and confidential are not the same word
This distinction is routinely blurred, and blurring it is how organisations lose trust they cannot easily recover.
- Anonymous means nobody — including the survey administrator — can connect a response to a person. No identifiers are collected, and reminders cannot be targeted.
- Confidential means responses are linked to individuals but access is restricted and reporting is aggregated. This is what almost every “anonymous” corporate survey actually is.
Confidential is a legitimate and often better choice: it allows follow-up, longitudinal tracking and non-response analysis. What is not legitimate is calling it anonymous. If a survey sends personalised reminders to non-respondents, it is not anonymous, and employees notice.
Say which one it is, in one sentence, at the top. Precision here buys more honesty than any reassurance does.
What people are actually protecting
Programmes tend to assume the fear is of formal consequence — discipline, a note on a file. It is usually something softer and more rational: being seen as negative, as not coping, or as the person who is behind.
Transformation frequently carries implied insecurity, and the effects are documented. The meta-analysis by Magnus Sverke, Johnny Hellgren and Katharina Näswall in the Journal of Occupational Health Psychology found job insecurity has detrimental consequences for employees’ job attitudes, organisational attitudes, health, and to some extent their behavioural relationship with the organisation — with behavioural effects more pronounced among manual than non-manual workers.
An employee who is unsure whether their role survives the programme has a poor incentive to record that they cannot yet do the new process. Saying so is a disclosure about their own competence at the moment competence feels load-bearing.
The same pattern shows up wherever it has been measured directly. Microsoft and LinkedIn’s 2024 Work Trend Index, surveying 31,000 knowledge workers, found 52 per cent reluctant to admit using AI for their most important tasks and 53 per cent worried it would make them look replaceable. People do not misreport because they are dishonest. They misreport because the honest answer costs something.
The practical consequence: readiness surveys are biased optimistic in exactly the groups under most pressure. A function reporting high confidence and low capability elsewhere in your evidence is not contradicting itself; it is telling you the survey was not safe.
The segmentation trade-off, stated honestly
| Fewer identifiers | More identifiers | |
|---|---|---|
| Honesty | Higher, especially in small or exposed teams | Lower where the answer could be attributed |
| Actionability | Low — an organisational average hides everything | High — you can see which function is at risk |
| Follow-up | Impossible | Possible, which is often the point |
| Failure mode | True but useless | Precise and wrong |
There is no configuration that avoids the trade-off. The workable position is to collect the segmentation you genuinely need for decisions, and then bind yourself publicly on how it will be reported.
Publish the reporting rules before the survey opens
This is the intervention that does most of the work, and it costs nothing:
- A minimum reporting threshold. No group smaller than a stated number is reported separately — five is a common floor. State the number in the invitation, not in a footnote afterwards.
- Who sees what. Whether a line manager will receive their own team’s results, and at what size. This is the question respondents are actually asking.
- What free text does. Verbatim comments are the highest-risk field for identification. Say whether they will be quoted, paraphrased or read only in aggregate.
- What the data will not be used for. An explicit statement that responses will not inform individual performance assessment.
Suppression is not only an ethics measure; it protects the data. A team that knows its six responses will be shown to its manager will produce six agreeable responses, and the number you get back will be precise and wrong.
Response rate is a finding, not an administrative statistic
A 34 per cent response rate is usually reported as a caveat. It is better read as a result.
Low response is rarely about survey length alone. It generally means one of three things: people do not believe anything will happen, they do not feel safe answering, or they genuinely have no capacity that week — and the third is itself a readiness finding worth acting on.
Non-response is also not random. The people with least time and least confidence are the least likely to respond, which biases results in the reassuring direction. Comparing response rates by function is often more informative than comparing scores: the function that did not answer is telling you something.
What reliably improves both honesty and participation is not a better invitation email. It is having visibly acted on the previous round — naming two things that changed because of what people said. Surveys that lead to observable action attract answers; surveys that disappear into a report do not, and the effect compounds across waves.
More on readiness measurement in the Change Readiness Hub, or see how to design the instrument before worrying about the response rate.