Measuring Employee Satisfaction with Surveys: How to Do It

Guess what people in Britain blame most often for making them ill at work. Not noise and not heavy lifting: 964,000 workers were suffering from work-related stress, depression or anxiety in 2024/25, and the work factors they name most often are workload pressures and a lack of managerial support (Health and Safety Executive, annual statistics published 20 November 2025; cause data averaged over 2009/10 to 2011/12). Neither of those turns up in a personnel file.

You measure employee satisfaction by asking your team anonymously and voluntarily about six areas separately, so the job itself, the line manager, the team, working conditions, development and pay, scoring every statement on one agreement scale, then reporting results only for groups large enough that nobody can be singled out. That is done with a survey: with empirio.ai, an online survey tool made in Germany, EU hosting is included in the free plan. By the end of this article you have a questionnaire, one headline figure and a reporting rule that still holds in a team of eight.

📌 The short version

• One overall score cannot say which area dragged it down.
• Six areas are enough: job, manager, team, conditions, development, pay.
• No UK statute sets a minimum group size for reporting.
• No works council veto exists, yet safety consultation duties can apply.
• Response rates fall when nothing followed the last survey.

Asking your team about satisfaction: where the honest answers come from

Honest answers about satisfaction arrive in writing and anonymously, almost never any other way. In a one-to-one with your own line manager, the person who would have to be criticised is sitting across the desk, and whoever raises it carries the risk alone.

The conversation still has its place. It settles individual cases, digs for causes and decides what happens next. As a measurement it fails, because the answers are not comparable and because no manager can survey their own effect neutrally. An employee survey with a standard questionnaire does the opposite: everyone gets the same statements in the same order on the same scale.

The usual shortcut is one big question, “How satisfied are you overall?” That looks thrifty and is not. A single score gives you a number nobody can explain, because it never says which part of the job pulled it down, and a drop caused by pay reads exactly like a drop caused by a new rota. HSE puts the same limit on its own staff survey, describing it as “a broad indicator of the situation in an organisation” and “a starting point rather than giving a clear diagnosis” (Health and Safety Executive, Management Standards).

A workable survey carries both. The area questions show you where to start, and one overall question gives you a figure you can hold against next year.

What goes in the survey: which questions measure employee satisfaction

Employee satisfaction is measured with statements people agree or disagree with, covering six areas: the job itself, the line manager, the team, working conditions, development and pay. Each area gets three to five statements, so that one clumsily worded sentence cannot decide the score for a whole area on its own.

The six areas are not a free choice. HSE builds its own working conditions survey around six areas as well, namely demands, control, support, relationships, role and change, and support there covers the organisation, the line manager and colleagues alike. Leave an area out of your questionnaire and you still measure it, only silently: the dissatisfaction migrates into the free text box.

Wording decides the rest. A statement describes one observable thing, it carries no negative and it never asks about two things at once. “My line manager is fair and good at the job” is useless, because nobody can tell which half is being agreed with.

Six statements are a start and not yet a questionnaire, because a complete survey structure also includes an introduction, a block of demographic questions and a closing page. For a full draft, the guide to creating a questionnaire helps, because it also covers order, length and the pilot run.

Which scale suits satisfaction statements

Agreement scales with four to six points have become the norm for satisfaction statements, running from strongly agree to strongly disagree. An even number of points pushes everyone to one side or the other, an odd number allows genuine indecision, and the HSE analysis tool reports on a range of 1 to 5. Either choice is defensible; mixing them inside one block of statements is not. A separate metric question with its own scale at the end is another matter and does no harm. How such a scale is built and read is set out in the article on the Likert scale.

Open questions in a satisfaction survey: one will do

One open question at the end brings in the point no pre-written statement thought of. Keep it to one or two, because free text has to be read and sorted by hand and because it compares badly across years. The form that works is “What should we change first?”, since it asks for a priority instead of a list.

Metrics and analysis: what ends up in the report

A satisfaction survey ends with three numbers for every area: the mean, the share of agreeing answers and how many people answered. A first report needs nothing more than that, and without the third number the first two cannot be read at all.

The distribution behind the average explains why. Picture a hypothetical team answering one statement on a five-point scale, with the mean landing on 3.0: that figure turns up when everybody answers in the middle, and it turns up just as readily when half the team is delighted and the other half is drafting a resignation letter. The second case is the dangerous one and the mean hides it completely. So print the distribution next to every mean, or at least the share of the top two points.

Mean, top two box and eNPS: what each number does

The top two box is the share of answers on the two most agreeing points, so strongly agree and agree together. It stands up better than the mean, because it does not assume the gaps between points are equal, and it can be explained to anybody without a word of statistics. Alongside it, many teams use the employee Net Promoter Score for a single headline figure, which asks on a scale of 0 to 10 how likely someone is to recommend the employer. That score measures willingness to recommend rather than satisfaction, and it only says something once you have several measurements to hold against each other. Work it out with the eNPS calculator.

The employee satisfaction index: one figure out of several areas

An employee satisfaction index rolls the area scores into a single number. There is no binding formula behind it: some organisations take the unweighted mean across all statements, others weight the six areas by how much each one matters to their own workforce. Either is defensible, as long as the calculation is written down and stays the same from year to year. Skip that note and your second measurement quietly compares something other than the first, with nobody in the room able to say so. Index values from other employers fail as a benchmark for exactly the same reason, because you cannot see what went into them.

Indirect metrics such as staff turnover measure something else

Staff turnover, sickness absence and overtime get treated as an indirect measure of satisfaction. They are easy to get hold of and they say almost nothing about it: low turnover can mean loyalty or an empty job market, and high absence can come from workload or from a flu wave. As early warning signs these rates earn their place, as a satisfaction metric they do not. Report them alongside the survey and never instead of it.

A metric without its response count beside it is not a metric, it is a claim.

Anonymity, data protection and consultation before you send it

Three things need settling before a satisfaction survey goes out: the smallest group you will report on, which data your survey tool picks up alongside the answers, and whether anybody has to be consulted first. Those three decide whether staff experience the survey as anonymous, and the honesty of the answers follows from that.

Order matters. Promise anonymity first and work out the reporting rules afterwards, and you are left with a choice between breaking the promise and publishing a report too coarse to act on. Both get noticed.

How small a group can you report on

No UK statute sets a minimum group size for reporting employee survey results. Neither the UK GDPR nor the Data Protection Act 2018 names a figure, so the test is the outcome: whether anybody can still be picked out of the numbers. The Information Commissioner’s Office calls that risk singling out, which it defines as being able to “isolate the records that relate to a person from a dataset”, and it asks you to weigh the “means reasonably likely to be used” to identify someone (ICO, anonymisation guidance).

Context does the work here, which is why a fixed number travels badly between organisations. The ICO offers the example that a family name “may be enough to distinguish them from others in the context of their workplace, but not in the context of the general population”, and it describes identifiability as sitting “on a spectrum”. An average score for a team of four lives in exactly that territory. Set a floor before you send, name it in the invitation, and roll small units together in the report.

When an anonymous survey is not anonymous in law

Answers stay personal data for as long as anything ties them back to a person. The Data Protection Act 2018 defines personal data as “any information relating to an identified or identifiable living individual”, and somebody counts as identifiable where they can be identified “directly or indirectly” (Data Protection Act 2018, section 3). The usual culprits are personalised invitation links, a login, a block on answering twice, a stored IP address and raw data sitting with a supplier. Consent is also the wrong instinct for anything identifiable, because the ICO holds that consent “is not usually appropriate in the employment context, due to the imbalance of power between you and your workers”. Say in the privacy notice which of those apply, before the first question. How a survey becomes anonymous in technical terms is covered in the article on the anonymous survey, and the legal side in the one on survey data protection.

Do you have to consult staff or their representatives first

No general statutory duty makes you consult anybody before running a voluntary satisfaction survey in the UK, and there is no works council holding a veto of the German kind. The Information and Consultation of Employees Regulations 2004 come closest and work differently: they apply to undertakings with at least 50 employees, and the duty to negotiate an agreement only starts once at least 10 per cent of employees ask for one in writing, subject to a floor of 15 people and a ceiling of 2,500. What then has to be shared is the development of the business, the employment situation and decisions likely to lead to substantial changes in work organisation. A satisfaction survey is none of those.

Two duties can still bite, depending on what you do with the results. Run the survey as part of your stress risk assessment and employees who are not represented by safety representatives must be consulted “in good time on matters relating to their health and safety at work”, in particular on “the introduction of any measure at the workplace which may substantially affect” their health and safety (Health and Safety (Consultation with Employees) Regulations 1996, regulation 3). And the ICO expects an employer introducing monitoring to “seek and document the views of your workers or their representatives (such as trade unions), unless there is a good reason not to”; a voluntary anonymous survey is not monitoring, so read that as the regulator’s view of the employment relationship rather than a rule about surveys. Where a union is recognised, talk to it early whatever the law strictly requires. This article is not legal advice.

What a tool for an anonymous employee survey has to do

A tool for a satisfaction survey has to manage four things: let people take part without an account and without a personalised link, carry blocks of statements on one shared scale, count the answers area by area, and come with a written processing contract. Miss the fourth and the survey cannot run in your organisation at all, however good the questionnaire is.

That fourth point is law, not procurement etiquette. Under Article 28(3) of the UK GDPR, “Processing by a processor shall be governed by a contract or other legal act under domestic law, that is binding on the processor with regard to the controller”, and Article 28(1) lets a controller “use only processors providing sufficient guarantees” that the right technical and organisational measures are in place (UK GDPR, Article 28). In a staff survey you are the controller and the survey provider is the processor, so that contract is yours to have signed before the first invitation goes out, not afterwards.

Check what a plan actually covers before you commit to it, and check the plan you really intend to use. With empirio.ai, EU hosting is included in the free plan, the data processing agreement is included in the free plan, sharing by link is free and the QR code generator is free, which matters for sites where nobody sits at a screen (as of October 2026). The limits belong in the same breath: access restriction with anonymous answers starts from Plus, and cross tabulation and filters start from Plus as well, which is exactly what you need to break results down area by area. If you want to try the questionnaire on one team first, you can create a free online survey and stay inside the free plan with 50 participants a month.

Why sceptical teams do not answer

Doubt about anonymity is the most common reason for a thin response rate, and no reassuring sentence talks it away. What works is technology people can check for themselves. So before you invite anyone, find out what your tool demonstrably does not collect: no login, no personalised link, no block on answering twice, no IP address in the raw data. Put those four points into the invitation word for word and name your minimum group size in the same paragraph. Write “of course it is all anonymous” instead and you get exactly the cautious answers you advertised for.

Two more lines save a lot of argument later. Say in the invitation who can open the raw data, and set down that free text is read through for identifying detail before any of it reaches a line manager. A questionnaire to try out takes a few minutes to put together once you start with a free account.

Sensible goals and real risks: an employee survey without the let-down

A satisfaction survey delivers a stocktake and a running order, not a solution. It shows which of the six areas scores worst and how far it sits from the rest. What to do about it appears in no answer.

Disappointment nearly always grows out of the wrong expectation of the instrument. Start a survey in order to raise satisfaction and you have the mechanism the wrong way round, because satisfaction rises through what you change afterwards. The survey itself moves one thing only, the expectation that something will happen.

What a satisfaction survey does for you

• It makes problems visible before the first resignation arrives.
• It compares areas against each other and across the years.
• It reaches everyone at once, including the quiet ones.
• It shows processes from the view of the people using them daily.

What you take on in return

• Responses drop when nothing followed the last survey.
• Small teams cannot be reported on separately.
• Free text costs reading time and compares badly.
• A single score with no previous year says almost nothing.

Satisfaction is neither engagement nor performance

Job satisfaction, as this article uses the term, is an evaluative attitude towards the work situation, so a judgement rather than a behaviour. Engagement means something else: William Kahn examined in 1990 in the Academy of Management Journal the conditions under which people put themselves into their work role, and named three of them, meaningfulness, safety and availability. Somebody satisfied is therefore not automatically engaged, and somebody working with real commitment is not automatically satisfied. A satisfaction survey measures the attitude and not the behaviour, and performance it measures least of all.

From results to an action plan

A satisfaction survey turns into an action plan in four steps: put the figures on the table, pick one single topic, put a name against it and set a date to check back. Everything beyond that is decoration, and it usually just delays the first step.

Speed beats thoroughness here. A workforce that has heard nothing four weeks after the survey already has its answer, just not the one you meant to give.

Poor results: the first two weeks

A weak score is information and not an attack, however it feels on the morning you read it. So share it in full and without polish, including the areas that came out badly. Hold figures back and you confirm the very suspicion that pushed the score down in the first place. A plain format does the job: the score for each area, last time’s score beside it, and two sentences underneath on how you read it.

One topic a quarter instead of ten actions

Ten actions on one slide look decisive and never get done. Take the worst scoring area instead, look in the free text for the reason people give most often, and derive exactly one change from it that can be finished inside a quarter. The next topic comes the quarter after. After a year four things are done, and that is more than most action plans deliver.

How staff spot a token gesture

Any action that does not touch the point people raised reads as a token gesture. Answering criticism of workload with a fruit bowl therefore lands worse than doing nothing at all: it shows the message was read and not understood. Just as recognisable is the action with no date and no name against it. Say openly what you cannot change and give the reason for it. An honest no to a pay demand costs less trust than a promise that fails to appear next quarter.

One point of law reaches into the plan itself. Where the survey is part of your stress risk assessment, the consultation duty described earlier applies to what you do next and not only to the questionnaire: employees who are not represented by safety representatives must be consulted in good time on measures at the workplace that may substantially affect their health and safety (Health and Safety (Consultation with Employees) Regulations 1996, regulation 3). In practice, put the draft plan in front of people before you sign it off rather than announcing it finished. Where a union is recognised, bring it in at the same stage.

Common mistakes in employee satisfaction surveys

The four mistakes below cost no money and plenty of trust, and three of them only surface with the next survey. They turn up regardless of size, in a village club exactly as in a department of 200.

One cause runs through all four. The survey gets planned as a data collection rather than as a promise. Send the questionnaire and you have also undertaken to read the answers, to treat them confidentially and to do something with them.

The survey starts with no plan for the results

The most expensive mistake happens before the first question. Fail to agree who sees the results when and who decides on actions, and you land in a question of ownership after the analysis, while the report sits around for three months. Staff read that silence as the answer, and the next round loses responses. So fix a date for the feedback session before you invite anyone, and name it in the invitation.

One overall question is asked to cover everything

A single question about overall satisfaction is quick to analyse and leads nowhere. It hands you a figure that cannot be explained, because it never says which area is pulling it down. Six areas with three statements each barely add to the analysis and tell you where to begin.

Outside benchmarks instead of your own last figure

Sector figures for satisfaction are tempting and rarely comparable. They come from other questionnaires, other scales and other workforces, and often the year of collection is missing altogether. The only comparison that genuinely fits you is your own from the last survey. Which is why the first measurement is worth running even though its result means little on its own.

Small teams get reported separately

A report that breaks out a team of three gives its members away, even with no name beside the figure. The person concerned knows immediately, and the manager does the arithmetic. A report like that echoes for two years, because the next survey gets answered more carefully. Roll small units together, even where a question of detail then goes unanswered.

Conclusion: the employee satisfaction survey, and what comes after

Measure six areas roughly rather than one overall figure precisely. A survey of 20 statements, one open question and a clear rule on group size gets you further than any polished instrument that never goes out. And fix the date of the feedback session before you share the link, because the value of your first measurement only appears with the second.

What to read next

• Ready to work through the answers? Evaluating a questionnaire
• Unsure when a question should be open and when closed? Open and closed questions
• Want to understand the recommendation question? Net Promoter Score

Ready for the first survey?

With empirio.ai you create the survey anonymously, send it out by link and see the results area by area without building a spreadsheet.

Create an employee survey with empirio.ai