Proxy Metrics

A long time ago, way back when I still had hair, Microsoft introduced Team Foundation Server and one of the productivity metrics commonly reported on was lines of code written per developer. At the time, it seemed like a relevant metric. More code meant more output, no?

Pretty quickly, we all viewed that metric with a healthy degree of scepticism. Anyone with the right exposure knows that some of the best engineering work involves removing code, simplifying systems, and solving problems with fewer lines rather than more. The developer who writes the least code is sometimes the one delivering the most value.

The issue isn’t that lines of code is useless. It's just a poor predictor of the thing we actually care about: valuable software landing in the hands of customers.

Decades later, we risk making the same mistake, only with Jira or GitHub instead of TFS. We’ve swapped lines of code with Jira tickets closed, story points completed, sprint velocity, deployment frequency, and countless other delivery metrics. These can provide useful signals, but they are still proxies. They are measures of activity, not necessarily measures of outcomes.

This is where Goodhart’s Law becomes relevant:

“When a measure becomes a target, it ceases to be a good measure.”

As soon as people know they are being measured against a metric, they begin optimising for the metric. That’s not a criticism, it’s human nature. It’s also why the old saying, “what gets measured gets done”, remains true.

The challenge for leaders is deciding what to measure in the first place.

For product and engineering teams, I’ve increasingly come to believe that we should measure outcomes, not activity.

These are not outcomes:

A story point
A Jira ticket
A pull request

These ARE outcomes

A customer adopting a feature.
A reduction in churn.
An increase in conversion, retention, revenue, or profitability.

Those measures are still imperfect, but they are much closer to the value the business is trying to create.

Over the years, I’ve looked for the perfect productivity metric. I still haven’t found one.

What I have found is that high-performing organisations tend to have two things in common.

They focus relentlessly on outcomes rather than activity.
They build cultures of trust.

Hire people you trust.

Then trust the people you hire.

But trust doesn’t mean blind faith. It means trust, but verify. Metrics still matter. The role of metrics is to provide signals, identify trends, and help leaders ask better questions. They should inform judgement, not replace it.

If you don’t trust your people, no dashboard is going to solve the problem.

If you trust your people but have no visibility into outcomes, that’s a problem too.

I’m curious where others have landed on this. Have you found metrics that genuinely correlate with value creation, or are we all still searching for the least imperfect proxy?