Just play around with it quickly and a few thoughts about it. Experiments metrics are often compare against other experiments. Currently the metrics are plot as Metric vs Version. * Version name does not has sufficient metadata to tell what is that experiment exactly (I think there are no obvious way to name the version in a custom way, correct me if I am wrong) * Slicing/Filtering is a must for experiment tracking. i.e. Compare experiment_001 with a new feature and experiment_002 without a new feature. * I used kedro viz locally, however, experiment metrics are most likely more useful with a Server. It may works fine if it is a 1 person project, but as long as multiple person are involved, there will be need of a central server to host all metric datasets to have a complete history (i.e. tracking progress of a particular project)