Hi Eleanor,
Test coverage is what I’d need the most. It would let me see at a glance which projects don’t have enough tests, and public shame the owning team 
Now, we could imagine many other metrics worth monitoring: the time to perform a specific step in the pipeline, the size of the artifacts, …
Here is a proposal on how this could work:
I believe we could leverage the meta-data functionality to define builds custom metrics. Something like that:
# pipeline.yaml
steps:
- label: Run tests
command: run_tests.sh
statistics:
- label: Code coverage
unit: percent
meta-data: code_coverage
# run_tests.sh
<perform tests + collect coverage metric>
buildkite-agent meta-data set "code_coverage" "$COVERAGE"