Proxy Metrics and Reality Gap

~390 Words | ~1.5min Read

Metrics are a wildly powerful tool for leaders. Used well, they enable teams to incrementally improve their process even when the leader isn’t in the room. But to be useful, metrics must be timely and obviously connected to the team’s actions. “Did we make a profit last month?” isn’t usually either of those for most teams. So we measure proxies instead.

Most metrics are proxies. They’re never the actual thing you care about. You don’t care about velocity or story points; you care about delivering customer value. You don’t care about tickets closed; you care about solving customer problems. This gap between what you measure and what you actually want creates space for gaming.

Consider a mechanic shop measuring how many times mechanics wipe down their tools. The outcome we want is clean tools that last longer. But if the rag is full of oil, wiping doesn’t make tools clean—it just spreads oil around. We’re measuring an activity correlated with the outcome, not the outcome itself. Or consider sales, and cold calling. Making calls is correlated with getting new client. But a thousand cold calls with zero closures isn’t the behavior we want to encourage.

The same pattern shows up in software. A developer solving the most bugs sounds great until you realize he’s also introducing more bugs. We don’t actually want him resolving more bugs; we want him driving product quality up. Bugs are just a proxy for quality.

So why use proxies at all? Because delayed feedback wastes months of potential improvement. By the time you know if you made a profit last quarter, you’ve lost one or two months of possible course correction. Proxies give us earlier visibility and faster feedback than waiting for the actual outcome. The cold call today connects to the contract three months from now. We need leading indicators even when they’re imperfect.

Proxies are a necessary evil. But they can be managed if we remain cognizant of the kind of proxy we’re using and how to keep our systems in balance. The proxy is not the thing itself. Used well, proxies help teams see what matters and improve what they can’t yet directly measure. The question isn’t whether to use proxies. The question is: do you know what kind of proxy you’re using, and what balancing metric keeps your system honest?