How to run an employee engagement survey

An engagement survey is a cheap way to find out what a company's people think, and an expensive way to teach them that nobody is listening. Which one you get depends almost entirely on what happens in the four weeks after the survey closes. This guide covers the whole loop for a company between 20 and 300 people, using a free form tool and a spreadsheet.

Decide the questions, then cut half of them

Keep the survey to 15 to 25 questions, each a statement rated on a five-point agree-to-disagree scale, plus two open text boxes. Cover a fixed set of themes so you can compare over time:

Role clarity: I know what is expected of me. I have what I need to do my job.
Manager: My manager gives me useful feedback. My manager treats people fairly.
Growth: I can see a path to grow here. I have learned something new in the last six months.
Recognition: Good work is noticed. I feel valued for what I do.
Workload and wellbeing: My workload is sustainable. I can take leave when I need to.
Leadership and direction: I understand where the company is going. Leaders communicate honestly.
Overall: I would recommend this company as a place to work. I see myself here in two years.

The two "overall" questions are your headline numbers. The two open questions should be specific: "What is one thing we should change" and "What is one thing we should keep". Vague prompts like "any other comments" produce vague answers.

Avoid double-barrelled statements ("My manager is supportive and gives clear direction") because the respondent cannot disagree with half of them. Avoid questions you are not willing to act on; asking about pay satisfaction when the increment budget is fixed only generates a complaint you have already decided not to address.

Anonymity that people believe

Anonymity is a promise about behaviour as much as technology. Three rules keep it credible:

A reporting threshold of five. Never show results for any group smaller than five respondents. A team of four cannot be reported separately; combine it with its neighbour. This is the single rule that makes people in a small company answer honestly.
Collect only the demographics you will use. Department and tenure band are usually enough. Asking for gender, location and level in a 40-person company makes every response identifiable.
Say who sees raw text. Open answers should be read by one or two named people, lightly edited to remove identifying detail, and only then shared with managers as themes.

If you are using a general-purpose form tool, turn off response tracking and email collection, and say so in the invitation.

Running it

Give the survey a two-week window, send it on a Tuesday, and have the founder or CEO send the invitation personally with one line on why it matters and what happened after the last one. Send two reminders. Aim for at least 70 percent participation; below that, the results reflect the people with the strongest feelings in either direction. Managers should block 15 minutes in a team meeting for people to fill it in, which removes the "I was too busy" problem without making anyone feel watched.

Run it at most twice a year. A quarterly survey in a small company produces survey fatigue and no time to act between rounds.

Reading the results

For each statement, compute the favourable score: the percentage of respondents who agreed or strongly agreed. Averages hide the shape of the answers; 60 percent favourable with a large neutral middle is a different problem from 60 percent favourable with a large unfavourable tail.

Then look at three things. The lowest-scoring themes overall. The themes where one department scores far below the others, which usually points at a manager or a workload problem. And the change from the previous survey, which matters more than the absolute level once you have two data points. Read every open-text answer; in a small company this is a few hours and it is where the real information is.

Resist the urge to benchmark against published industry scores. Question wording, timing and culture differ too much for the comparison to mean anything. Your own last survey is the only benchmark worth having.

Acting on it: the part that decides everything

Within two weeks of closing, share the headline results with everyone. Show the favourable scores, name the three lowest themes without spin, and quote a handful of anonymised open comments including critical ones. Then commit to two or three specific actions, each with an owner and a date. Not ten. Two or three that will visibly happen.

Each manager gets their team's results, with the threshold rule applied, and holds one team discussion on them. The manager's job is to listen and pick one team-level action, not to defend the scores.

Ninety days later, send a one-paragraph update on each committed action. This single message is what makes the next survey's participation go up instead of down. Where the action is a policy change, write it into the handbook and announce it on whatever internal channel people actually read, such as the company feed.

What not to do

Do not run a survey in the month after a layoff, a missed increment or a controversial policy change and expect it to tell you anything beyond that. Do not ask managers to guess who wrote which comment. Do not let a low score for one manager become a performance case without a conversation and a second data point. And do not skip the survey because you expect the results to be bad; that is precisely when people most need to see that someone asked.