Skip to content

Destination ranks

Destination-rank scoring evaluates whether the model ranks a worker’s observed next employer highly among time-valid sampled alternatives. It reports HCT and lagged firm-size ranks globally and within destination-defined industry and global-year size-band diagnostic pools.

This task needs exported company and mover embeddings plus prior-year observed company employment counts. It is intended for next-employer prediction evidence, not for producing the standard observed-match score dataset.

The scorer is a model CUDA task. Run and resume instructions, inputs, output grain, and downstream use are documented in Destination ranks.