How Meta predicts tech attrition: ML model results surprised the researcher

A People Analytics specialist with over a decade of experience at Meta decided to investigate why talented professionals leave tech companies in their first year of work. He was confident in his answer — until the moment he launched a Machine Learning model to analyze the data.

Initial

Hypothesis After years of working with HR data, the analyst had formed a firm belief: employee turnover is determined by two or three key factors. Most likely, he reasoned, these would be salary, career growth, and onboarding quality. Conversations with colleagues at Meta confirmed this view. But when it came to collecting concrete data and building a model, the picture became less clear.

The

Model Revealed Something Different The ML model discovered that the predictive power of the factors he was betting on was far more modest than expected. Instead, unexpected patterns emerged as the focus. It turned out that those who often leave are not the most dissatisfied with their salary, but those who felt uncertainty about where their career was heading within the company.

The study identified the true predictors: Employee career trajectory prior to current role Speed of decision-making at the manager level Cultural signals a person receives in the first few weeks Motivational alignment between personal goals and company mission * Presence of a mentor or sponsor in the early stages ## How to Apply This The research opens a new direction for HR strategy. Rather than focusing solely on compensation and career ladders, companies can invest in onboarding quality, assign a sponsor to each new employee, and accelerate decision-making processes at the team lead level.

"Data doesn't lie — the problem is often not what you see on the surface"

What

It Means Tech companies spend enormous resources attracting talent, but lose them due to shortcomings in the integration stage. The ML model shows: if the first few months are organized correctly and a person is given a clear direction, many can be retained. This is a tool for People Analytics teams to rethink their approach to employee retention.