How to run an exit interview survey

To run an exit interview survey that people answer honestly, send it from the people team after the resignation is settled and at least a week before the last day, say exactly who reads it and what reaches the manager, ask what started the search and whether anything could have changed the outcome, and let the leaver decide at the end how much may be shared. Then summarise the answers on a fixed cycle and act on the themes, because a survey that changes nothing teaches the next leaver to write “better opportunity” and move on.

The reason to bother is in the numbers. Gallup reports that 52 percent of voluntarily exiting employees say their manager or organisation could have done something to prevent them leaving, and 51 percent say nobody spoke with them about their job satisfaction or future in the three months before they left. Gallup puts the cost of replacing an employee at one half to two times their annual salary, and calls that conservative.

Timing: between the resignation and the last day

The window is narrow. Before the resignation is accepted the person is still negotiating, and on the last afternoon they are returning equipment and saying goodbye. The useful time is the stretch in between, when the decision is made and the relationship is still civil.

Send the link from the people team, not the line manager, within a couple of days of the resignation being accepted. The exit interview form from Pellow and Vance asks for the last day as a date, which tells the readers how much time remains for a conversation if the answers call for one. One reminder a week later is enough. More than that and the survey becomes one more thing the organisation wants from someone who is leaving.

Keep the survey apart from the formal steps. Resignation, notice and final pay have their own rules, which the Fair Work Ombudsman sets out, and nothing about the survey should suggest any of them depends on it.

Anonymity: promise confidential, not anonymous

Anonymous means nobody can tell who answered. Confidential means somebody can, and you are promising what they will do with it. Almost every exit survey is the second kind.

Look at what the example asks first: team, length of service in bands, and last day. In a firm of a few hundred, that combination points to one person most weeks. The example is honest about this. Its welcome screen says the answers are read by two people in the people team, do not go on the personnel file, do not change any reference and do not reach the manager unless the leaver says so. It never claims to be anonymous. Its last question offers “Yes, all of it, with my name on it” even though there is no name field, because the people team can match a response to a person from the team and the date. That is exactly why the promise has to be confidentiality.

Two practical controls help. Limit who can open the form, and limit who sees responses. A stored response holds the answers, the form version and the time it was submitted, and nothing about the respondent unless a question asks for it. If you add an email alert or a webhook, remember that the full answers go wherever that alert or webhook points, so send them to a named reader, not a shared inbox.

In Australia, the OAIC notes that a private sector employer’s handling of employee records is exempt from the Privacy Act when it is directly related to the employment relationship. That exemption does not make careless handling a good idea. The Fair Work Ombudsman’s workplace privacy guide describes best practice employers as having clear policies on what they collect, what they keep and when it can be passed on. Write that policy for exit answers before the first survey goes out. None of this is legal advice.

The question mix

Good exit surveys ask about the start of the decision, the pull of the new role, and the moment it could have gone another way. The example does all three in fourteen questions.

What started it. A single choice of ten reasons, including “I was not looking, this came to me”, with a description asking for the thing that started it rather than the thing that finished it. A second, optional multiple choice collects everything else that contributed.

What made them accept. A required open question, which the example says is the one the quarterly summary is built from.

A short agreement grid. Six statements, from having the tools needed to being able to raise a concern without it costing anything. Agree and disagree scales attract acquiescence bias, where people agree regardless of content, so keep the grid short and the statements specific. The statements closely mirror those in an engagement survey question bank, which lets the people team compare leavers against the staff who stay.

The moment. “Was there a moment when something could have changed your mind?” is optional, and its description says it is worth two minutes. It is the only answer that can help someone still employed.

Keep the tone of the whole survey consistent with a pulse survey or a 360 feedback form: specific statements, a not seen answer where someone could not have observed the thing, and open questions that stay optional.

Who reads it

Say who they are, in the survey. The example says its answers are read by two people in the people team, in its intro and again on the welcome screen, so nobody has to hunt for the promise.

Then let the leaver decide what travels. The final question offers three levels of sharing: everything with a name, the substance without a name, or nothing beyond the people team. Asked last, once the answers are written, it gets more generous choices than asked first. Readers pass on only what was agreed, in the form chosen. A manager who wants to know “who said that” gets the same answer every time.

Exit answers about a manager are one input, not a verdict. Read them next to the manager’s own performance review history and the team’s pulse results before drawing a conclusion.

Turning answers into action

Harvard Business Review’s April 2016 article on exit interviews opens with a department of 17 people where, a year after a new manager arrived, only eight remained. Four had resigned, and all four exit interviews told the same story about missing leadership skills. The deeper finding was that the organisation promoted managers for technical skill rather than managerial skill, and the executive committee changed the promotion process.

That is the model. Read each response within a week and record two things: the first reason and the moment answer. Every quarter, count first reasons by team, suppressing any team where so few people left that a count would identify someone. Pick the recurring themes, agree one change for each with a named owner, and tell current staff what changed. The example’s thank-you screen promises exactly this, a quarterly summary alongside everybody else who left.

The process, step by step

The table at the end of this article lays out ten steps from writing the reading policy to telling current staff what changed, with the timing and the owner of each. The first two steps happen once. The rest repeat for every leaver or every quarter.

Common mistakes

Sending it from the manager. The manager may be the reason, and the leaver knows the manager will see the sender line.

Calling it anonymous. When a response is later matched to a person, every future leaver hears about it.

Asking only for ratings. A grid tells you something is wrong. The open questions tell you what.

Reading responses once a year. By then the leaver’s team has lost two more people.

Collecting reasons and never reporting back. Staff who see nothing change stop believing the survey is read.

Build it

The example is a conversational form, one question per screen, built from dropdowns, a date, single and multiple choice fields, long text and a rating matrix, which is a small share of the 41 component types available. Descriptions show under each question on a conversational form, which is where the example explains why each question matters.

For access, the login required page explains limiting a form to signed in people, to a list of addresses or to everyone on a company domain; it suits staff who already have an account. Where leavers do not, a secret key sent privately by the people team is the practical alternative. The access control guide walks through all four access types. Add the email alerts add-on only if it points at one of the named readers, and use a thank-you screen to repeat what happens next. For question wording beyond this survey, see how to write survey questions.