Employee check-in questions that tell you who is about to leave

One question does most of the work in an employee check-in: "Looking at the next twelve months, how do you feel about staying on here?" Ask it early and plainly. Let the answer decide what you ask next. If someone sounds settled, ask what keeps them. If they hedge, ask what is holding them back and whether they have told anyone. If they say they might leave, ask what is driving it and what one thing would change their mind.

Then ask everyone two more questions. One is about their manager, and one is about whether training prepared them. Close with one number: how likely they are to recommend working here, from 0 to 10. That is ten questions in five to ten minutes. It is the script behind the check-in calls we run.

The rest of this guide covers the full question set and the reason for each question. It also covers the part most check-in templates leave out: turning the answers into something HR can act on the same week.

There are three reasons, and they add up.

The questions are closed. "On a scale of 1 to 5, how satisfied are you with your onboarding?" gets a 4 from almost everyone. A 4 is the polite answer, and nobody wants to explain a 2. You end up with a spread of numbers and no reasons.

The wrong person asks. When the manager runs the check-in, nobody can talk about the manager. On the calls we run, the supervisor is one of the most common reasons people give for wanting to leave. They do not say it to the supervisor.

Nobody acts on it. If a check-in ends in a spreadsheet nobody opens, the employee learns not to bother next time. The answers have to reach someone who can act, with the person's name attached. Otherwise the check-in is wasted.

"Looking at the next twelve months, how do you feel about staying on here?"

It works because it asks about the future instead of about satisfaction. Someone can be unhappy and staying, or happy and already interviewing elsewhere. A satisfaction question tells you the mood. The twelve-month question tells you the decision.

Read the whole answer, not just the first word. "Yeah, I think so, but let's see how the appraisal goes" is not a yes. It is unsure, and unsure people are the group you can still help. Liking the work is also different from planning to stay. Score what they plan to do.

Do not read options aloud. If you ask "would you say you are committed, unsure, or considering leaving?", people pick the last word they heard. Or they pick the one that sounds least dramatic. Ask the question open, and sort the answer into a group afterward.

"What are the main things that make you want to stay and build something here?" This asks what keeps them. It is a different question from what might push them out, which is covered below. Never assume someone is happy with a thing just because they did not complain about it. If someone has not mentioned their manager, they have not told you their manager is good.

"What should we make sure we keep doing, so it stays this way for you?" These answers are the cheapest retention plan you will ever get, because they are things you already do. They are also the things that quietly stop when a team gets busy.

"If you had to name the one thing holding you back, what would it be?" Ask for one thing only. The biggest concern is the one worth acting on. Asking for just one keeps the answer from turning vague.

"Have you been able to raise that with anyone here?" This is often the most useful answer in the whole record. If nobody at work has heard the concern, that is a communication problem you can fix tomorrow. If it was raised and ignored, that is a management problem. It also tells you the employee has already tried.

"What would need to change for you to feel more settled about staying?" Ask for a specific change. You want something a manager can put in a calendar.

"What is the main thing making you think about moving on?" Again, ask for the main thing. In almost every group, the answers fall into a short list: a better offer or no growth here, a role that does not match what was promised, pay, the supervisor, work culture or hours, and personal reasons.

"Have you been able to raise that with anyone here?" Ask it again here, because the answer changes what happens next. If they never raised it, the concern goes to the manager's manager. If they raised it and were ignored, it goes higher.

"If one thing could change and make you stay, what would it be?" You will not be able to grant every request. But you will be surprised how often the request is small. It might be a talk about growth, a clearer target, a move to another team, or a pay review that was promised and never happened.

"How satisfied were you with the support and guidance from your manager during this period?" Ask everyone, including the people who are settled. That way your data on each manager does not come only from unhappy people.

"How effective was the training in helping you do your job?" Score the answer afterward on a five-point scale, from very effective to not effective. Training that did not prepare people is one of the easiest causes of early exits to fix. It is also one of the least reported, because nobody wants to admit they did not understand the product.

"On a scale of 0 to 10, how likely are you to recommend working here to a friend?" Ask it last, after the person has said what they needed to say. If you ask it first, it colors every answer after it.

The number is most useful next to the other answers. Someone who says they are staying but would not recommend the place is not low risk. They are medium risk, and they are telling you something the twelve-month question could not.

Open answers are honest, but you cannot add them up. Score every call into the same fields, using fixed lists. Keep the person's own words next to each category.

Commitment: committed, unsure, or at risk, with one sentence that sums up what they said. Risk drivers (reasons they might leave): one main reason and up to two more, from the fixed list above. Stay drivers (reasons they stay): up to three, from a separate list, only for people who were asked. Requested actions: what they asked for, from a list such as more one-on-ones, clearer goals, skill training, a move inside the company, a fairer workload, recognition, or a pay review. Raised with anyone: yes, no, or not asked. Training effectiveness on the five-point scale. Recommend score, 0 to 10.

Then use a simple rule to set a risk level, instead of a model. High: said they may not stay, or raised something that needs a formal review, such as harassment. Medium: unsure, or settled but would not recommend. Low: settled and would recommend. Unknown: never reached, or declined. Never count unknown as low. Someone you could not reach has not been measured, so you cannot assume they are fine.

A rule works better than a model score because you can explain it in a meeting. "The model said 71" ends a retention program at its first review. "She said she may not stay, and here is the sentence" starts a real conversation.

Not the manager, for the reason above. Ideally not HR either, if HR sits in the same building as the employee. A neutral voice gets a different answer. A phone call also gets a different answer from a form, because people say things they would not type.

Tell people up front that the check-in is not anonymous. You cannot act on an answer with no name. So say it before the first question: what you share goes to senior HR leadership. If someone objects, end the call and never call them again. An anonymous check-in program feels safer, but it gives you nothing you can act on.

Run it in the language the person actually speaks, in plain everyday words. Keep it short. Five to ten minutes is long enough to get past the polite answer. It is also short enough that a branch employee will take the call.

With more than a handful of employees, this has to be a calling program instead of a series of meetings. That is why we built AI check-in calls. The AI counselor phones each person at the milestone you set and runs this script. The answers land in a list with names on it.

In order, with the follow-up branch noted. Everyone gets the opening, the gate question, the two questions for everyone, and the close.