Employee Health Score FAQs

Employee Health is predictive signal technology designed to assist People teams and managers understand which employees might be at risk of leaving the organization. It surfaces early indicators of disengagement so HR leaders and managers can take timely, informed action to retain employees. 

What does the model look at? 

Employee Health looks at anonymized patterns in how employees interact with Lattice over time. Specifically, it considers:

• Recent changes in their manager or title
• How long the employee has been at the company
• How frequently the employee has given and received feedback or praise
• The sentiment of their written updates
• Manager type (individual contributor, manager, etc.)

These signals are based primarily on usage patterns over the preceding 12 months and reflect behavioural trends that may correlate with disengagement or upcoming departure.

What kind of model is used? 

Employee Health uses a statistical method called logistic regression. This allows it to combine multiple signals into a single risk score that estimates the likelihood an employee might leave in a given month. The model has been trained on years of historical data across existing customers and is regularly evaluated for fairness and accuracy.

Does the model use personal or sensitive information? 

Employee Health does not use any sensitive or protected characteristics. That includes:

• Age
• Race or ethnicity
• Gender or sexual orientation

Lattice is committed to ethical AI and ensures our model only uses signals related to employee behaviour within Lattice, not personal demographics.

Is this model making decisions about employees? 

No. The model is intended to be a supporting signal, not a decision-making tool. It is designed to help HR teams and managers focus their attention where it might be needed most, so they can ask questions, offer support, and strengthen employee engagement before attrition becomes a reality. 

What do the health statuses mean? 

The mathematical breakdown of high health, medium health, and low health can be summarized in the table below:

Note that, if a user is labeled as high health it does not automatically mean that user will stay at the company. It simply means that, based on historical employee data, we predict this user has a greater probability of staying at the company than a low health employee.

What do I need to do to make sure Employee Health works best?

There are two ways to ensure Employee Health works best at your company:

• Regular usage of the Lattice toolset (regular updates, 1:1s, feedback, etc.) will yield more accurate predictions.
• Maintaining accurate employee information (title changes, organizational hierarchy, accurate start and end dates, etc.) will enhance the model’s accuracy at your organization. 

How should I use Employee Health insights?

Lattice recommend's using Employee Health as one input in a broader people strategy. Pair it with what you know from conversations, engagement surveys, and team dynamics. The model is there to provide an additional input to aid—not to replace—human judgment.