Model drift

What is model drift?

Model drift is a decline or change in an AI model's performance over time, as the data it sees or the world it models moves away from what it learned. A lead-scoring model trained before a price increase, a new service line or a downturn can keep producing confident scores that no longer match reality. Drift is gradual, so it rarely shows up as an error message.

Why it matters to a revenue leader

Revenue leaders already know drift from forecasting: stage weightings that worked two years ago slowly stop matching how deals close. AI models drift the same way. NIST's AI Risk Management Framework names data, model and concept drift as reasons AI systems need ongoing monitoring and maintenance. In a buy-and-build platform, every add-on brings new customers, service lines and sellers, which changes the data a model sees. Someone has to own checking that the model still earns its place.

A fire protection company with 20 sellers uses a model to rank which inspection customers are most likely to approve repair quotes. In its first two quarters, 70 percent of top-ranked quotes close. After the company adds a large national account with a slower approval process, that rate falls to 45 percent over the next two quarters. Nobody touched the model. The customer mix moved underneath it.