THE ISSUE

The stay interview is the most promising retention instrument HR has adopted in a decade. Ask people why they stay while they are still here, instead of only asking why they left once nothing can be changed. The logic is sound. The instrument is cheap. And in most organizations it has exactly one channel, and that channel runs through the direct manager.

Which means the one thing the instrument cannot reliably detect is a manager. Or the environment that manager is responsible for.

You cannot train your way out of wiring. And the stay interview is the only instrument in the standard HR toolkit that reaches employees while they still have something to protect. That is a lot to lose to a default nobody chose.

WHAT THE RESEARCH SAYS

Let’s start with what we actually know about stay interviews, which is less than the practitioner literature suggests. The published base consists largely of quality improvement projects and conference reports rather than controlled studies, and it rests heavily on self-reported satisfaction with the conversation rather than on turnover outcomes. The strongest quantitative finding on retention outcomes is a moderate association between receiving a stay interview and remaining employed, in a sample of 132 direct support professionals at a single provider (Taylor-Cook, Spreat, Hicks & Wilson-Fall, 2024).

Two things in that literature deserve more attention than they get.

The first is a design claim. Freel et al. (2023) describe the stay interview as a conversation held entirely separately from performance evaluation. That is a statement about how the instrument is intended, not a measurement of how employees experience it. Nobody has tested whether the separation holds in the mind of the person being asked.

The second is an open question. McCoy et al. (2025) list, among the ethical implications organizations must resolve before implementing stay interviews, who should conduct them and how sensitive disclosures should be handled. The field has raised the question. It has not answered it. In the meantime, the implementation guides that do exist, including the 14-question guide developed by Freel’s group, put the direct manager in the chair.

The research on upward communication tells us what that default costs.

Milliken, Morrison and Hewlin (2003) interviewed 40 employees. Around 85 percent described situations in which they had withheld a concern from a supervisor. The most frequently named reason was fear of being labeled negatively and damaging the relationship. Among the withheld topics were the ones an organization would most need: a supervisor’s competence, fairness and ethics problems, harassment and abuse. And 74 percent believed colleagues were staying silent about the same things.

Silence of this kind is not diffuse. It is aimed. The meta-analytic record places leadership at the center of employee silence (Hao et al., 2022), and Kiewitz et al. (2016) identify the mechanism: fear mediates the path from abusive supervision to defensive silence, amplified by a perceived climate of fear. The listener determines which topics are sayable at all.

Then there is what happens to people who do speak. Burris (2012) ran three studies, including one with 281 restaurant managers, separating challenging voice from supportive voice. Employees who challenged existing practice received lower performance ratings, were rated less loyal, and had their ideas adopted less often. Employees who defended existing practice were rated better. The stay interview asks for challenging voice. The evaluation system prices it.

Wu et al. (2023) go one step further: employee voice can trigger identity threat in the leader, and through that mechanism produce abusive supervision. Speaking up does not only fail to be heard. Under some conditions it generates the behavior it was reporting.

Fast, Burris and Bartel (2014) close the loop. Managers with low managerial self-efficacy solicited less input, reacted more negatively to employees who spoke, and adopted fewer suggestions. The mechanism was ego defensiveness. The manager least able to hear the answer is the manager most motivated not to ask the question. The completion rates will look identical either way.

IN PRACTICE

A pattern that repeats across sectors. An organization rolls out stay interviews after a bad turnover year. Managers get a question guide, a target of two conversations per employee per year, and a completion rate in the HR dashboard.

Twelve months later the completion rate is 94 percent. The aggregated themes are compensation, career path, and workload. No unit stands out. The department with the highest voluntary exits, the highest sick leave, and the most vague, never-formalized complaints returns a completely unremarkable set of stay interview notes.

Nothing was falsified. Every conversation happened. The data are accurate. The measurement is not.

The notes record what people were willing to say to the person who writes their performance review. The instrument worked as wired. It measured what employees will tell their manager about their manager, and the organization read that as a measurement of itself.

THE STRUCTURAL QUESTION

Here the argument has to survive a serious objection, and it is worth stating in full before answering it.

The objection: the manager is not the flaw, the manager is the mechanism. Tangirala and Ramanujam (2012) show that consultation by one’s own manager raises upward voice, mediated by the employee’s sense of having influence. Detert, Burris, Harrison and Martin (2013) show that voice improves unit performance when it reaches the leader with authority to act, and damages it when it flows to people who have none. Liu, Tangirala and Ramanujam (2013) show employees already route voice by relationship quality, going to the skip-level leader when the direct relationship is weak. And a conference paper by Xu, Liu and Jia (2012) reports that skip-level voice costs the person who uses it, in salary growth and in citizenship behavior, precisely when the direct manager holds high authority over the issue raised. That last finding was never published in a peer-reviewed journal, so treat it as a signal rather than a settled result. It points the same way as Burris.

That objection is correct, and it kills the obvious fix. Taking the manager out moves the information to someone with less power to change the unit, and charges the employee for the detour.

So the design question is about the channel, not the chair. Why does an instrument built to detect problems in a unit have only one, and why does it terminate at the person it would have to report on?

Four decisions follow.

One. Add a second channel, do not swap the first. The manager conversation stays, because the relational effect is real and documented. Alongside it, a second route that does not terminate in the reporting line: skip-level, ombuds, HR outside the chain, works council. Two channels, two datasets, and a comparison between them. Where the two diverge sharply in a unit, that divergence is itself the finding.

Two. Treat the output as unit-level data. Stay interviews are filed as individual retention notes. Their diagnostic value is comparative: this unit against that one, this quarter against the last. Organizations already hold the indicators, and the chain between them is well established. Destructive leadership correlates most strongly with counterproductive work behavior, the largest effect in the meta-analytic record (Schyns & Schilling, 2013). And counterproductive work behavior aggregates: measured at unit level it tracks with turnover, productivity and customer satisfaction (Carpenter, Whitman & Amrhein, 2020). What is missing is not the metric. It is the join between the conversation and the metric.

Three. Read absence as a signal. Uniformly bland responses in a unit with elevated turnover are information, not a null result. So are late scheduling, cancelled conversations, and completion rates that arrive suspiciously close to 100 percent. Most reporting systems record these as good performance.

Four. Price the risk honestly, including the risk of the second channel. An organization that asks people to name what would drive them out, without a mechanism preventing the named person from learning who said it, has built a list rather than a diagnostic. And an organization that opens a second channel without protecting the people who use it has simply relocated the cost. The EU Whistleblower Directive settles this principle, but only for a defined catalogue of reportable breaches. Member States may extend that catalogue, and Austria has. Even so, almost none of it reaches the everyday conduct through which toxic leadership actually works.

ONE TAKEAWAY

If your stay interview data shows no problem units, you have learned nothing about your organization. You have learned something about your instrument.

There is a fast test. Take the unit with your highest voluntary turnover and read its stay interview notes. If they are unremarkable, the question is not why people are leaving a unit where everything is fine. The question is who was sitting across the table.

Have you ever been asked what would make you stay, by the exact person who was the reason you were considering leaving? I would like to hear how that conversation went.

Markus

This essay first appeared on 26 August 2026 in my LinkedIn newsletter Toxic by Design.

---

Sources

Burris, E. R. (2012). The Risks and Rewards of Speaking Up: Managerial Responses to Employee Voice. Academy of Management Journal, 55(4), 851-875. DOI: 10.5465/amj.2010.0562

Carpenter, N. C., Whitman, D. S. & Amrhein, R. (2020). Unit-Level Counterproductive Work Behavior: A Conceptual Review and Quantitative Summary. Journal of Management, 47(6), 1498-1527. DOI: 10.1177/0149206320978812

Detert, J. R., Burris, E. R., Harrison, D. A. & Martin, S. R. (2013). Voice Flows to and around Leaders: Understanding When Units Are Helped or Hurt by Employee Voice. Administrative Science Quarterly, 58(4), 624-668. DOI: 10.1177/0001839213510151

Fast, N. J., Burris, E. R. & Bartel, C. A. (2014). Managing to Stay in the Dark: Managerial Self-Efficacy, Ego Defensiveness, and the Aversion to Employee Voice. Academy of Management Journal, 57(4), 1013-1034. DOI: 10.5465/amj.2012.0393

Freel, S. A., Helm, S. L., Hanes, L., Lee-Chavarria, D., Hogentogler, R. E., Brock, A. et al. (2023). Stay Interviews: Guiding Meaningful Conversations for Retention. Journal of Clinical and Translational Science, 7(s1), 33-34. Conference abstract. DOI: 10.1017/cts.2023.197

Hao, L., Zhu, H., He, Y., Duan, J. & Zhao, F. (2022). When Is Silence Golden? A Meta-analysis on Antecedents and Outcomes of Employee Silence. Journal of Business and Psychology, 37(5), 1039-1063. DOI: 10.1007/s10869-021-09788-7

Kiewitz, C., Restubog, S. L. D., Shoss, M. K., Garcia, P. R. J. M. & Tang, R. L. (2016). Suffering in silence: Investigating the role of fear in the relationship between abusive supervision and defensive silence. Journal of Applied Psychology, 101(5), 731-742. DOI: 10.1037/apl0000074

Liu, W., Tangirala, S. & Ramanujam, R. (2013). The relational antecedents of voice targeted at different leaders. Journal of Applied Psychology, 98(5), 841-851. DOI: 10.1037/a0032913

McCoy, M., Cannedy, S., Oishi, K., Canelo, I., Hamilton, A. B. & Olmos-Ochoa, T. T. (2025). Using Stay Interviews as a Quality Improvement Tool for Healthcare Workforce Retention. Quality Management in Health Care, 35(1), 38-43. DOI: 10.1097/qmh.0000000000000465

Milliken, F. J., Morrison, E. W. & Hewlin, P. F. (2003). An Exploratory Study of Employee Silence: Issues that Employees Don’t Communicate Upward and Why. Journal of Management Studies, 40(6), 1453-1476. DOI: 10.1111/1467-6486.00387

Schyns, B. & Schilling, J. (2013). How bad are the effects of bad leaders? A meta-analysis of destructive leadership and its outcomes. The Leadership Quarterly, 24(1), 138-158. DOI: 10.1016/j.leaqua.2013.07.001

Tangirala, S. & Ramanujam, R. (2012). Ask and You Shall Hear (But Not Always): Examining the Relationship Between Manager Consultation and Employee Voice. Personnel Psychology, 65(2), 251-282. DOI: 10.1111/j.1744-6570.2012.01248.x

Taylor-Cook, M., Spreat, S., Hicks, D. & Wilson-Fall, T. (2024). Does the Use of Stay Interviews on a Regular Basis Enhance Direct Support Professionals’ Retention? Social Innovations Journal, 28, Article 8971.

Wu, L., Long, L., Hu, S. & Xu, Y. (2023). An identity threat perspective on why and when employee voice brings abusive supervision. Frontiers in Psychology, 14, 1133480. DOI: 10.3389/fpsyg.2023.1133480

Xu, E., Liu, W. & Jia, L. (2012). The cost to go the extra mile: Consequences of skip-level voice. Academy of Management Proceedings, 2012(1), 13525. Conference paper, not peer-reviewed. DOI: 10.5465/ambpp.2012.13525abstract