Ask The Analyst: Attrition models can flag dozens of flight risks at once – what separates organizations that act effectively from those that drown in the output?

Each week for The Briefing, UNLEASH’s weekly intelligence email for senior business decision-makers, we ask the true experts – our community of analysts – to solve the biggest HR challenges.

Got a question for the analysts? You can up it directly to them via The Briefing – make sure you’re signed up.

This week’s question: Attrition models can flag dozens of flight risks at once – what separates organizations that act effectively from those that drown in the output?

Here are their perspectives.

Rebecca Wettemann, CEO & Principal Analyst, Valoir:

Organizations that act effectively on attrition model signals follow a few common best practices including:

They correlate probability with causation. They pair red flags with driver-level context, like compensation or a dysfunctional manager, so HR or managers know what they should address to prevent attrition.

They understand that not all attrition is bad. They tier red flags by not just risk level but the business impact of the loss of that person and react accordingly based on role scarcity, succession depth, cost to replace, or other business-critical factors.

They audit and refine models and signals over time. They track when employees actually leave despite an intervention, which interventions work, and what signals don’t appear to be valid, to avoid signal fatigue.

Finally, and perhaps most importantly, their models align with their organization’s capacity to act. A ton of red flags that just become part of a reporting cycle only add to stress and negativity. Effective models are triage systems aligned with an organization’s ability to act – whether it’s in terms of manager or HR bandwidth or budget for counter-offers or other engagement investments – so flight risks can be prioritized and addressed in a way that minimizes risk to the business.

Michael Rochelle, Chief Strategy Officer & Principal Analyst, Brandon Hall Group:

The difference is not the quality of the prediction; it is the quality of the talent operating model that surrounds it. Across our research and case studies, one finding emerges consistently: organizations that achieve measurable talent outcomes do not treat AI predictions as answers. They treat them as inputs into a disciplined decision-making process.

Technology identifies where to look; leadership determines what to do next. Organizations that struggle with predictive attrition often fall into what we call the "list trap." Their models produce a growing list of employees with elevated flight risk, but managers receive little context, few decision criteria, and no clear guidance on how to respond.

The result is predictable: inconsistent interventions, competing priorities, and eventually, inaction. High-performing organizations take a fundamentally different approach. They have built an operating system that converts prediction into action:

• They establish "governance before intervention”.
• They "prioritize business risk instead of reacting to individual risk scores".
• They develop "repeatable intervention playbooks.
• They recognize that "manager capability is the true differentiator".
• They create continuous learning loops.

Taken together, these examples point to a broader conclusion. The organizations that succeed with predictive attrition are not distinguished by having more advanced AI. They are distinguished by having a more mature decision architecture.

Cheney Hamilton, Research Director, Bloor Research:

Honestly, it’s the wrong question. Attrition models, and the exit surveys behind them, ask why someone is leaving, whether it's salary, whether we can save them.

But the evidence consistently suggests that even successfully retained employees tend to leave within six to twelve months anyway. The decision was made long before the flag appeared; retention efforts just delay the inevitable. So, the organizations that "act effectively" aren't acting on the flight risk at all, they're acting on the role.

Here's why that matters: 90% of job descriptions don't reflect the work someone actually performs. When a person leaves, the real role, the invisible work, the relationships, the judgement, walks out the door with them.

The business then recruits against a fictional job description, so the next hire is the wrong hire, can't hit the ground running, inherits invisible work nobody scoped, and looks worse for it. They leave too. That's how one departure becomes a chain of attrition that corrodes culture, attitudes and behavior.

The effective organizations use the exit moment to map the real role before it disappears, so the next hire is hired into the job that actually exists.

Nicole Roberts, HR Industry Analyst, 3Sixty Insights:

Attrition models generate signals, not answers. A high flight-risk score tells you where to look, but it doesn't tell you why someone is at risk or what action will actually make a difference.

The organizations that get this right have a plan before they ever turn the model on. They know who owns the response, which risks require action, and how they'll measure whether those actions actually improve retention. Otherwise, it's just another dashboard that creates alert fatigue.

They also recognize that not all flight risks are created equal. The goal isn't to retain everyone. It's to focus your time and resources on the people and capabilities that are most critical to delivering the business strategy.

If losing someone won't materially impact the business, they shouldn't receive the same level of intervention as a critical role or high-impact performer.

The most mature organizations also look beyond individual scores. If flight risk consistently shows up around a particular manager, function, or career stage, that's a signal to improve the system – not just react to the individual.