A stay conversation is often counted as complete when the meeting happens. That count answers a scheduling question, not a retention question. The useful question is whether the conversation produced a documented action and whether the employee saw that action occur. Separating completion from follow-through changes what an organization measures, what it can claim, and what it should fix.
Published November 3, 2026
Define the three layers of evidence
A stay conversation program produces three kinds of evidence. The first is participation: who was offered a conversation and who attended. The second is content: what concerns were raised and what actions were agreed. The third is follow-through: whether the agreed action happened and whether the employee observed it. Most programs measure the first layer and assume the other two follow.
Each layer requires a different source. Participation comes from calendars and program records. Content comes from notes or a structured form. Follow-through comes from the action owner's records and a check with the employee. A study that uses only one source can describe one layer but cannot support a conclusion about the others. The report should name which layer each finding comes from.
Measure participation with a clear denominator
Participation is the easiest layer to measure and the easiest to distort. Decide whether the denominator is all employees, all employees in a target group, or all employees who were eligible during the period. Decide whether a canceled and rescheduled conversation counts once or twice. Decide whether a conversation initiated by the employee counts the same as one initiated by the manager. Write the rules down before counting.
Report participation by group, not just overall. A high overall rate can hide a team where no one was offered a conversation, or a shift that was systematically excluded. If the program is meant to reach everyone, the group breakdown is the meaningful result. If it is targeted, the target group should be defined in advance so the rate has a stable denominator.
Capture content in a comparable form
Free-text notes are rich and hard to aggregate. A short structured form is less rich and easier to compare. The best approach is usually a hybrid: a few structured fields for the concern category, the agreed action, the owner, and the due date, plus a small free-text space for context. The structured fields support the review, and the free text preserves what the categories miss.
Keep the form short enough that a manager will complete it honestly. A long form invites rushed entries and copied language. A short form with clear definitions produces better data, even if it captures less nuance. Review the definitions with the people who complete the form so the categories mean the same thing across teams.
Separate the action from the conversation
An action is not an action unless it has an owner and a date. A note that says "discuss workload" is a topic, not a commitment. A note that says the supervisor will review the shift allocation with the employee by a specific date is a commitment that can be checked. Require the owner and date before the conversation is marked complete.
This rule changes behavior. Managers begin to make smaller, more concrete commitments instead of broad promises. Employees get a clearer sense of what will happen next. The program becomes a series of verifiable steps rather than a listening exercise. The discipline is simple, but it requires the organization to treat the follow-up as part of the conversation rather than an optional extra.
Check follow-through independently
Follow-through should be checked against the action owner's records and, where possible, the employee's experience. The owner's record shows whether the action was completed. The employee's report shows whether the completion was visible and useful. Both matter, and they can disagree. An action can be completed administratively without changing anything the employee experiences.
Sample a set of actions each quarter and trace them from the note to the record to the employee check. Record the result for each step: completed on time, completed late, not completed, or no longer relevant. This trace is more informative than a completion percentage because it shows where the process breaks. It also gives the program owner specific cases to discuss with managers.
Treat self-report as a measure with limits
Employee surveys and interview responses are valuable because they capture perception. They are weak as measures of process. An employee may report that a concern was addressed when the underlying issue persists, or report that nothing happened when a change occurred outside their view. Neither response is dishonest; each reflects a different vantage point.
Use self-report to describe experience, and use records to describe process. When the two disagree, investigate rather than choosing the more convenient source. A disagreement may reveal a communication gap, a timing issue, or a genuine failure. The investigation is often more useful than the original metric.
Avoid measuring satisfaction as the only outcome
A stay conversation can be pleasant and change nothing. If the program is evaluated only by satisfaction scores, it will optimize for pleasant conversations. Include a measure of action completion and a measure of whether the employee's stated concern improved. Those measures are harder to collect, but they reflect the purpose of the program.
At the same time, do not dismiss satisfaction. A conversation that feels unsafe or perfunctory will not produce honest content, and the program will fail regardless of its action tracking. Satisfaction and follow-through answer different questions. The right approach is to measure both experience and follow-through and to treat them as related but distinct.
Protect confidentiality in a small sample
In a small team, a detailed action log can identify the employee who raised a concern. That risk is real and should shape what is recorded and who can see it. Store the content at the level needed to manage the action, and report only aggregates outside the immediate process. If a finding would identify an individual, suppress it or describe it in a way that protects the source.
Managers should understand the difference between managing an action and discussing a concern. The action can be discussed with the people who need to deliver it. The employee's identity and exact words should be shared only on a need-to-know basis. A breach of that boundary can end the program faster than any process failure.
Build a quarterly review
Review the program each quarter across the three layers. Report participation, the number of actions agreed, the share completed on time, and the employee-reported visibility of the change. Add a short qualitative note on the concerns that recurred. Compare the results with early attrition in the same groups, without claiming causation.
The review should end with one or two changes to the program itself. Maybe the form is too long, the action owners are unclear, or the follow-up window is unrealistic. Fix the program before asking managers to try harder. A quarterly rhythm keeps the program from becoming a ritual and gives it a chance to improve.
State what the program cannot prove
A stay conversation program cannot prove that it reduced turnover. Employees who stay may have stayed anyway, and employees who leave may have left for reasons the program could not address. The program can show that conversations happened, that actions were agreed, and that some actions were completed and observed. Those are meaningful outputs, but they are not causal proof.
The honest claim is bounded: the organization created a process for hearing concerns and following up, and the evidence shows how well that process ran. If attrition improved at the same time, that is a promising signal worth investigating. It is not a guarantee, and presenting it as one would undermine the credibility of the next review. It is also enough to justify continuing the program while the evidence accumulates.
Data Sources and Methodology
- BLS Job Openings and Labor Turnover Survey
- BLS Employee Tenure News Release
- DOL Fair Labor Standards Act
- Equal Employment Opportunity Commission
- NIOSH Total Worker Health
- NIST Privacy Framework
- Census Bureau Quarterly Workforce Indicators
- U.S. Office of Personnel Management
- U.S. Small Business Administration
- U.S. Department of Health and Human Services