How to Identify Flight Risk Employees: A Practical Scoring Model for HR Teams
Quick Answer
You can build a useful flight risk list in a spreadsheet using data HR already holds. Score each employee on a handful of signals (time since last promotion, manager change, team exits, pay position, blocked internal moves, pulse drops), rate the business impact of losing them separately, then sort into Red, Amber, and Green tiers. Use the tiers to decide who gets a conversation first, not to judge anyone. Back-test the model against your last 12 to 24 months of exits so the weights reflect your organisation, and review the list weekly.
What You Need Before You Start
Your list of employees with tenure, role, manager, team, last promotion or pay revision date, and pay position relative to range
Exit data for the last 12 to 24 months, with last working dates
The latest pulse or engagement data, if you have it
Two hours and a spreadsheet
You do not need software. You do need to be disciplined about the guardrails in Step 7.
Step 1: Define a Regretted Exit
Decide what counts as a loss you wanted to prevent. A simple definition works: a voluntary exit of someone who was rated at or above expectations, or who held a critical skill, or who would take more than a few months to replace. Everything in this model is aimed at those exits, not at all attrition.
Step 2: Back-Test Before You Build
This step is what separates a useful model from a guess.
List every regretted exit from the last 12 to 24 months.
For each, rewind to six months before they left. Using only the information you would have had then, record which signals from Step 3 were present.
Count how often each signal appeared across the group.
Do the same for a sample of people who stayed, so you can see how often the signals appear without an exit following.
If "manager changed" shows up in most of your regretted exits but also in most of your stayers, it is a weak signal for you. If "no promotion in 18 months" shows up in many exits and few stayers, weight it up. Your own data beats any generic list.
Step 3: Choose Your Signals and Starting Weights
Use these as a starting point and adjust after the back-test.
Signal | Starting points |
In a tenure window where your own exits tend to cluster | 1 |
No promotion, role change, or pay revision in 18 months or more | 2 |
Manager changed in the last 6 months | 1 |
Two or more exits from the same team in the last 6 months | 2 |
Pay below the mid-point of range, or below market for the role | 2 |
Applied for an internal move and was declined or blocked | 2 |
Manager or growth score dropped meaningfully in the last pulse | 1 |
Sustained heavy workload (extended overtime or unused leave building up) | 1 |
Maximum score: 12.
Keep the signal list short. Every extra signal makes the model harder to explain and harder to trust.
Step 4: Rate Impact Separately
Risk and impact are different questions. Do not fold them into one number.
Impact | Definition |
High | Critical skill, top performer, or hard to replace within a few months |
Medium | Solid performer with meaningful role knowledge |
Low | Readily replaceable, or in a role with a deep bench |
Impact decides who goes first when capacity is tight. Risk decides who needs attention at all.
Step 5: Sort Into Tiers and Assign Actions
Tier | Score | Action | Deadline |
Red | 7 to 12 | Stay conversation with a neutral or skip-level listener. Agree a specific response plan with the manager and HR. | Conversation within 7 days. Plan within 14 days. |
Amber | 4 to 6 | Manager-led career and workload conversation. HR checks that it happened and what came out of it. | Within 30 days |
Green | 0 to 3 | Standard cadence of one-to-ones and pulses. | Ongoing |
For Red and Amber, sort by impact. A Red with High impact is your first call.
Worked example (hypothetical)
A senior analyst has been with you 22 months (in a tenure window where exits cluster: 1), has had no promotion or pay revision in 20 months (2), their manager changed four months ago (1), two teammates have left in six months (2), and their pay sits below the range mid-point (2). Score: 8. Red. They own a critical model, so impact is High. They go to the top of the list for a stay conversation this week.
Step 6: Run a 30-Minute Weekly Review
Keep it short and consistent.
Who attends: HR business partner and one HR lead. Bring in a manager only for their own team members.
Agenda: New Reds and Ambers, overdue conversations, actions due this week, outcomes from last week.
Output: Each Red has a named owner, a date, and a status.
Rule: No Red stays on the list without a recorded conversation and a plan.
Step 7: Put Guardrails in Place
This model touches personal data and can affect how people are treated, so these are not optional.
Use it to listen, not to punish. Never use a score to withhold opportunities, change assignments, or influence performance ratings.
Be open that stay conversations exist. Employees should know the organisation checks in proactively. Secrecy damages trust if it surfaces later.
Restrict access. Limit the list to the few people who need it. Do not circulate scores to managers as labels.
Exclude sensitive data. Do not use health information, personal circumstances, protected characteristics, or activity-tracking data.
Check for bias. Periodically compare tier distribution across gender, location, and level. If one group is consistently scored higher, investigate before acting.
Check data protection requirements. Make sure the use of personal data aligns with your policies and applicable law, including India's data protection rules, and get legal input where unsure.
Step 8: Measure Whether It Works
Track these every quarter and recalibrate.
Metric | What it tells you |
Share of regretted exits who were Red or Amber six months before leaving | Whether the model catches what it should |
Share of Reds who left anyway | Whether your response plans are working |
Share of flagged people who stayed and why | Which actions actually helped |
Time from Red flag to first conversation | Whether capacity is the bottleneck |
Percentage of Reds with a recorded plan within 14 days | Whether the process is being followed |
If many exits were Green beforehand, you are missing a signal. If many Reds turn out fine without intervention, a weight is too high.
Spreadsheet Columns to Use
Employee ID, Role, Team, Manager, Tenure, Months since last promotion or pay revision, Manager change in last 6 months (Y/N), Team exits in last 6 months, Pay position, Internal move declined (Y/N), Pulse change, Workload flag, Risk score, Tier, Impact, Owner, Conversation date, Plan, Status, Outcome.
Common Mistakes
Overweighting pay. Pay matters, but many exits happen where pay is competitive. Let your back-test decide.
Building a black box. If HR cannot explain why someone is Red, managers will not trust it.
Flagging without acting. A list with no conversations is a liability, not a tool.
Telling managers "this person is a flight risk." It can change how they treat the employee. Share the action needed, not the label.
Ignoring quiet high performers. Quiet people rarely trigger obvious signals. Run periodic stay conversations beyond the list.
Treating the score as the answer. It tells you who to talk to first. The conversation tells you what is actually going on.
Where Neutral Listening Fits
A score points to who needs attention. It cannot tell you why. That comes from a conversation the employee trusts enough to be honest in, which is often easier with someone outside the reporting line. Use our guide to stay interview questions for prompts, and our piece on what employees say versus what they mean for how to probe past the polite first answer.
At AceNgage, trained neutral listeners run stay conversations for your highest-priority employees and return the underlying drivers and recommended actions, so your Red list turns into retention plans that actually get executed.
FAQs on Identifying Flight Risk Employees
What is a flight risk employee?
A flight risk employee is someone with a higher likelihood of leaving the organisation voluntarily, based on signals such as stalled growth, a recent manager change, team exits, pay position, or declining engagement.
Can HR predict who will resign?
Not with certainty. A scoring model helps prioritise who to talk to first, but it cannot predict individual decisions. Treat it as a way to focus listening, then rely on the conversation to understand the situation.
What data should I avoid using?
Avoid health information, personal circumstances, protected characteristics, and activity-tracking data such as keystrokes or screenshots. These raise fairness and trust concerns and can damage the credibility of the whole programme.
How often should the list be updated?
Refresh signals monthly and review Reds weekly. Recalibrate the weights each quarter using your latest exit and retention data.
Should managers see the scores?
Generally no. Share the action needed rather than the label, because labelling an employee as a flight risk can change how they are treated. Give managers enough context to have the right conversation.
What do I do when someone is Red?
Hold a stay conversation within seven days, ideally with a neutral or skip-level listener, then agree a specific plan with the manager and HR within fourteen days. Record the outcome and follow up.
Comments