THE IDEA
Measuring the shadow, not the thing
Some of the most important things are the hardest to measure. How do you measure learning? Wellbeing? Innovation? Trust? Organisational health? Customer loyalty? These things matter enormously. They resist direct measurement entirely.
So we use proxies. A proxy measure is something measurable that we believe correlates with the thing we actually care about. Test scores as a proxy for learning. Employee satisfaction surveys as a proxy for engagement. Patent filings as a proxy for innovation. Net Promoter Score as a proxy for customer loyalty. The proxy isn’t the thing. It’s a shadow of the thing - easier to see, easier to count, and never quite the same shape.
Proxies are necessary and dangerous. Necessary because you can’t manage what you can’t see, and proxies make invisible things partially visible. Dangerous because the proxy is always a simplification, and simplifications leave things out. The gap between what the proxy measures and what you actually care about is where the trouble lives. As long as everyone remembers that the proxy is a proxy - an approximation, not the truth - it serves well. The moment people forget, and start treating the proxy as the thing itself, the measurement system begins to mislead.
IN PRACTICE
When the stand-in takes centre stage
A university uses research publication count as a proxy for academic quality. Reasonable enough - productive researchers tend to be good ones. But over time, the proxy becomes the goal. Academics publish more, shorter papers. They split one study into three publications. They choose safe, publishable topics over risky, important ones. The publication count goes up. Whether academic quality follows is a question nobody’s asking anymore, because the proxy has replaced the question.
A company uses customer satisfaction scores as a proxy for product quality. Support agents learn that their performance reviews depend on the score, so they start asking customers to rate them highly at the end of calls. Some offer discounts in exchange for good ratings. The satisfaction score rises. Actual satisfaction - the real experience of using the product - hasn’t changed. But the proxy has been optimised, and the proxy is all anyone looks at.
A parent uses school grades as a proxy for their child’s development. The child learns to optimise for grades - studying for tests, working the marking criteria, managing teacher impressions. Their grades are excellent. Their curiosity, independence, and love of learning - the things the grades were supposed to reflect - may or may not be developing. The proxy is visible and celebrated. The real thing is invisible and untracked.
WORKING WITH THIS
Remembering what the proxy stands for
The first discipline is naming your proxies as proxies. When you choose a measure, say out loud: “This is a proxy for X. It’s not X itself.” This simple act keeps the gap visible. It prevents the slow drift from “this measure correlates with what we care about” to “this measure IS what we care about.”
The second discipline is regularly checking the correlation. Does the proxy still track the thing it’s supposed to represent? Correlations break down over time, especially when people are rewarded for the proxy. Schedule periodic reviews where you ask: if we improved this number, would we actually be better? If the answer is uncertain, the proxy may have drifted.
Use multiple proxies for important things. No single number captures learning, wellbeing, or innovation. Three imperfect proxies that approach the same thing from different angles give a much better picture than one imperfect proxy that everyone fixates on. And maintain some qualitative assessment alongside the numbers. Talk to people. Observe what’s happening. The things that matter most often reveal themselves in conversation and observation, not in dashboards.
THE INSIGHT
The line to remember
Every proxy measure is a simplification of something complex. The danger isn’t using proxies - it’s forgetting that they’re proxies and treating the shadow as the thing.
RECOGNITION
When this is in play
You’re relying on proxy measures when the number on the dashboard doesn’t match the reality you experience. When improving the metric doesn’t seem to improve the situation. When people start arguing about the methodology of the measure rather than about the thing it’s supposed to represent. When the measure was chosen because it was easy to track rather than because it captured what matters. When someone asks “but is it actually better?” after being shown an improving metric - that instinct is the gap between the proxy and the reality making itself felt.