usage metrics

Usage metrics are like a report card showing how much a service was actually used. Think of a gym: every time you swipe your membership card, an entry is logged, and there's also a table showing how many people showed up to each trainer's class. Administrators need these numbers to decide which services to expand and which to cut back.

The tricky part comes when multiple services are used at once. If a gym has only one trainer, counting entries alone is enough. But once there are several trainers, you need to break down "how many people came to whose class" trainer by trainer to make meaningful comparisons. Breaking things down into fine-grained categories like this is a key part of designing usage metrics.

Another thing to watch out for: numbers that look like they share the same name can still mean different things if they're counted differently. Just because two fields have the same name doesn't mean you can add them together or treat them as identical — doing so can give you a wrong picture of actual usage.

The article reports that "the GitHub Copilot usage metrics API released a feature that reports third-party agent app activity broken down by individual agent." One point to watch for here: the newly added field user_initiated_interaction_count counts something different from the existing top-level field of the same name. Just because the names match doesn't mean the two figures should be added together or treated as equivalent.

Look for a usage or billing menu in the admin or account settings of an AI service or dev tool you use regularly. 1) Check how daily/monthly usage counts are broken down. 2) If you use multiple tools or features together, compare whether the numbers are split by item or lumped into one total. 3) If you find two similarly named fields, check the documentation for what exactly each one counts to see whether they actually represent different values.