How Should You Analyse Employee Engagement Survey Results?

Running an employee engagement survey gives you data. The important part is working out what that data is actually telling you.

It can be tempting to start with the lowest-scoring questions, identify the departments with the lowest results and read through the most negative comments. But this can produce misleading conclusions.

A lower score is not necessarily a meaningful difference. A handful of strongly worded comments may not represent the wider workforce. And an organisation-wide average can hide substantial differences between groups of employees.

Good employee engagement survey analysis therefore needs to go beyond simply ranking scores from highest to lowest. It should help you understand the overall picture, identify meaningful differences between groups, interpret employees’ comments without compromising anonymity and, ultimately, focus leaders on the findings that deserve their attention.

What Should You Look at First in Employee Engagement Survey Results?

Start with the overall picture before looking at individual teams or questions.

Depending on the survey methodology, this might include headline measures such as an Employee Engagement Index (EEI) and Employee Net Promoter Score (eNPS), alongside the results for individual survey questions.

These measures provide useful reference points. They can help leaders understand overall levels of engagement, monitor changes between surveys and, where appropriate, make meaningful comparisons.

However, no single engagement score can explain everything that is happening within an organisation.

Two organisations, or two departments within the same organisation, could achieve a similar overall engagement result while employees have very different experiences of management, communication, development, wellbeing or working conditions.

Headline measures are therefore a useful starting point, not the end of the analysis.

Should You Use Mean Scores or Percentage Positive?

Another important consideration is how individual survey questions are reported.

Many employee surveys report the percentage of positive responses. For example, on a five-point agreement scale, responses of 4 and 5 might be grouped together and reported as the percentage of employees who responded positively.

This is easy to understand, but it can hide useful information.

Consider two employees who both move from a score of 4 to 5 between annual surveys. Their views have become more positive, but if both scores are simply categorised as “positive”, the percentage-positive measure records no change.

The same problem occurs at the other end of the scale. A movement from 1 to 2 may represent an improvement, but both responses could still be grouped together as negative.

For this reason, Protostar advocates using the mean average score on the 1–5 scale when analysing individual survey questions.

The mean retains more of the variation within the responses and makes it easier to identify movement between surveys.

This is particularly valuable when tracking results year on year. An organisation may be improving or declining gradually before enough employees cross the threshold between “negative”, “neutral” and “positive” for a percentage-positive measure to show a noticeable change.

Percentage-positive results can have a place as a simple summary measure, but they shouldn’t replace the more detailed information contained within the original response scale.

When the objective is to understand change, preserving that information matters.

Don’t Simply Focus on the Lowest Scores

One of the easiest mistakes to make when reviewing employee survey results is to assume that the lowest score must be the highest priority.

It might be. But not necessarily.

Some questions will naturally attract higher or lower scores than others. The organisational context also matters, as does what the question is actually measuring.

It is usually more useful to consider several things together:

• What are the strongest and weakest results?
• How have mean scores changed since the previous survey?
• How do different parts of the organisation compare?
• Are apparent differences statistically significant?
• What are employees saying in their written comments?
• Are there consistent patterns across several related findings?

This produces a much more balanced picture than simply creating a list of the ten lowest-scoring questions.

Look for Statistically Significant Differences Between Groups

Breaking employee survey results down by relevant respondent groups can reveal findings that are hidden within an organisation-wide average.

For example, depending on how the survey has been designed and the demographic information collected, you might compare results between departments, locations, job levels or other appropriate groups.

However, there is an important distinction between a difference and a statistically significant difference.

Imagine one department scores a survey question slightly lower than another.

It is easy to look at the two averages and conclude that employees in one department have a substantially worse experience.

But some variation between groups will occur naturally, particularly where the number of respondents is relatively small.

Statistical significance testing helps establish whether an observed difference is sufficiently large, relative to the amount of data available, to warrant greater attention rather than simply being normal variation in survey responses.

For HR teams and leaders, you don’t necessarily need to understand the underlying statistical calculations.

The practical question is:

Is this difference meaningful enough that we should pay attention to it?

Highlighting statistically significant differences can help leaders concentrate on the variations that deserve investigation rather than reacting to every small movement in the data.

Why Sample Size Matters

This is also why survey results from smaller groups need to be interpreted carefully.

Suppose a small team has a noticeably lower score than the organisation overall. With relatively few respondents, the views of a small number of people can have a substantial effect on the mean.

The result may still be important, but it needs context.

Larger respondent groups generally provide more evidence on which to base comparisons, while results from smaller groups may be more susceptible to variation.

There is also a separate issue of confidentiality.

Breaking results into increasingly small groups might appear to provide greater detail, but it can increase the risk of employees feeling identifiable.

Good employee survey analysis therefore isn’t about segmenting the data as far as technically possible. It is about finding the level of analysis that provides useful organisational insight while maintaining appropriate confidentiality.

Use Dashboards to Explore the Results

Employee survey dashboards can make a large volume of data much easier to explore.

Rather than relying on a static organisation-wide report, authorised users can filter results by the respondent groups included within the survey design and examine different parts of the organisation in more detail.

This can help answer questions such as:

• Which areas are performing particularly well?
• Where are experiences different from the organisation overall?
• Are there particular questions where results vary between groups?
• How have mean scores changed since the previous survey?

However, access to a dashboard shouldn’t be confused with analysis.

A dashboard allows you to explore the data. The more important task is interpreting what those results mean and deciding which findings deserve attention.

Without that interpretation, it is easy to spend considerable time moving between filters and charts without reaching a clear conclusion.

Why Detailed Data Packs Still Matter

Dashboards are useful for exploring results interactively, but organisations may also need detailed data packs. These provide HR teams and other authorised stakeholders with a more comprehensive view of the survey findings and allow the results to be examined in greater depth.

Different audiences will also need different levels of information.

An HR team responsible for the survey may need detailed results. A senior leadership team is more likely to need a clear summary of the most important findings, significant differences and major themes.

Managers may need information relevant to their own area so that they can understand the employee experience locally and contribute to appropriate action.

Good reporting is therefore not about giving everybody every piece of data.

It is about making the right level of information available to the right people.

How Should You Analyse Free-Text Employee Comments?

Quantitative survey questions tell you how employees responded to the issues you asked about.

Free-text questions give employees the opportunity to explain their views in their own words and raise issues you may not have anticipated. This can make comments extremely valuable.

It can also make them difficult to analyse.

A survey may generate hundreds or thousands of comments. Reading them individually can make it difficult to distinguish recurring issues from isolated opinions. There is also a natural tendency for particularly strongly worded or memorable comments to attract disproportionate attention.

Instead, comments can be analysed and summarised into themes and sentiment. For example, comments might reveal recurring themes concerning communication, management, workload, development opportunities or working arrangements.

Looking at the sentiment associated with those themes can then help establish whether employees are discussing an issue positively, negatively or in a more mixed way.

The objective is not simply to count comments. It is to help leaders understand the main messages emerging from what employees have said.

Should You Use AI to Analyse Employee Survey Responses?

The rapid development of generative AI means it is increasingly tempting to upload employee comments to an AI tool and ask it to identify the main themes.

Organisations should be cautious about doing this.

The first consideration is privacy and data protection.

Employee survey comments can contain sensitive information. Employees may mention colleagues, managers, health issues, workplace incidents or other personal circumstances. Before placing employee feedback into any external AI system, an organisation needs to understand exactly how that information will be processed, stored and protected.

There is also a more fundamental issue: employee comments often require human interpretation.

Language in employee surveys can be ambiguous. Employees use irony, shorthand and organisation-specific terminology. A comment may appear negative when read in isolation but have a different meaning when considered alongside the question being answered or the wider organisational context.

Several comments may also describe the same underlying issue in completely different ways.

AI can be useful as an analytical aid, particularly when dealing with large volumes of text, but it should not be assumed that automated analysis removes the need for experienced human judgement.

For employee survey feedback, accuracy is not simply a matter of categorising words as positive or negative. The analyst needs to understand what employees are actually trying to communicate.

That combination of privacy concerns and the need for contextual interpretation means organisations should think carefully before relying on AI alone to analyse employee comments.

Why Not Just Give Managers All the Comments?

There is another important reason for summarising free-text feedback: anonymity.

Employees can sometimes inadvertently make themselves identifiable through what they write.

A comment might mention a particular incident, job role, location, manager or set of circumstances that makes it possible for someone familiar with the situation to work out who probably wrote it.

Simply distributing raw comments widely can therefore undermine the confidentiality employees were expecting when they completed the survey.

Summarising comments into themes and sentiment allows the organisation to benefit from employees’ qualitative feedback while reducing the risk of individuals being identified. It also keeps attention on the collective message rather than on trying to determine who wrote a particular comment.

That distinction matters.

Employees need confidence that a survey is intended to understand their experiences, not identify individual respondents.

Bring the Numbers and Comments Together

Quantitative and qualitative findings are most useful when considered together.

For example, the mean scores might show that employees have concerns about communication from senior leaders.

The free-text analysis might then identify recurring themes around the timing of communications, uncertainty during organisational change or employees wanting more explanation behind important decisions.

The numbers have identified an issue.

The comments have provided additional context.

Equally, free-text comments may raise a recurring issue that wasn’t adequately covered by the structured survey questions. That could be useful when deciding what requires further investigation or when designing the next survey.

Neither source of information needs to stand alone. Together, they can provide a much richer understanding of employees’ experiences.

Should Every Difference Lead to an Action?

No.

One of the purposes of good analysis is to prevent that from happening.

A comprehensive employee engagement survey can generate a large number of findings. Attempting to create an action for every lower score, significant difference and negative theme can quickly produce an unmanageable improvement plan.

Analysis should instead help leaders identify what is genuinely important.

That may include issues affecting substantial numbers of employees, statistically significant differences between groups, recurring themes in written feedback, changes since previous surveys and areas that are particularly important in the organisation’s current circumstances.

There will also be strengths worth understanding.

If one part of the organisation consistently achieves particularly positive results, it may be useful to explore what is happening there and whether anything can be learned or shared elsewhere.

Employee survey analysis shouldn’t only be a search for problems.

The Aim Is Insight, Not More Data

Modern survey technology can make enormous quantities of employee data available very quickly.

That doesn’t automatically make the organisation better informed.

The real value comes from turning the data into a manageable explanation of what employees are experiencing and where leaders should focus their attention.

That means using measures that preserve useful detail, such as mean scores on a 1–5 scale, alongside headline measures such as EEI and eNPS. It means looking for statistically significant differences rather than reacting to every apparent variation. And it means treating employees’ written comments with the confidentiality and human judgement they deserve.

Dashboards and data packs make the information accessible.

Analysis makes it useful.

In Summary

When analysing employee engagement survey results:

1. Start with the overall picture, including headline measures such as EEI and eNPS.
2. Use mean scores on the 1–5 response scale to retain variation and make changes between surveys easier to identify.
3. Don’t rely solely on percentage-positive scores, which can conceal movement within positive or negative response categories.
4. Examine individual question results rather than relying on a single engagement measure.
5. Compare relevant respondent groups, but don’t assume every numerical difference is meaningful.
6. Use statistical significance to help identify differences that genuinely warrant attention.
7. Use dashboards and detailed data packs to explore the findings at an appropriate level.
8. Analyse free-text comments into themes and sentiment rather than allowing individual comments to dominate the interpretation.
9. Protect employee anonymity and consider privacy carefully, particularly before using external AI tools to process employee comments.
10. Retain human judgement when interpreting qualitative feedback, where context and nuance can materially affect meaning.
11. Bring the quantitative and qualitative findings together to identify the main messages.
12. Don’t try to act on everything. Use the analysis to decide what deserves further investigation or action.

A good employee engagement survey should ultimately make a complex organisation easier to understand.

The objective isn’t to produce the greatest number of charts, tables or metrics. It is to give leaders reliable evidence about employees’ experiences so they can make better-informed decisions about what happens next.

Need Help Analysing Your Employee Survey Results?

Protostar provides fully managed employee engagement surveys, from questionnaire design and survey administration through to analysis, reporting and support with action planning.

Our reporting can include EEI and eNPS measures, detailed data packs, filterable dashboards, mean-score analysis, identification of statistically significant differences and carefully summarised themes and sentiment from employee comments.

This helps leaders focus on the findings that matter while maintaining appropriate confidentiality and ensuring the results are interpreted in context.

Learn more about our Employee Survey Services or contact Protostar for an informal, no-obligation conversation about your requirements.