Latency or something else as IGP metric?

If you run a global network with a backbone connecting all sites around the world, what would you use as the best “value” for your IGP metric?

Currently we use latency in ms multiplied by 10 to give us a metric to assign to OSPF. Eg a link between locations with 60ms of latency would be assign a metric of 600.

I don’t mind this approach as it generally results in traffic taking the lowest latency path, but it can become annoying to manage as latency can change depending on the circuit type (eg wavelength vs L2 service) and is also a bit tricky to automate the config for

So curious what others do for large global backbones