Focused Guide

Direct evidence versus translated signals

3 minute read

You may have plenty of data but little clarity. Some figures feel busy and loud. Others feel sparse but honest. The useful split is not complex. It is whether the evidence shows the goal itself or asks you to translate.

Direct evidence lets you read progress at a glance. If the goal is “publish clear articles,” a count of articles ready to publish can be read without a story. If the goal is “reduce defects,” confirmed defects per release is legible. Translated signals need you to add meaning. Hours spent, meetings attended, and drafts created can help, yet none states the outcome on its own. When you have to attach a because to make a measure relevant, you are doing interpretation work that can drown the goal.

Example: finished tasks beat hours logged

Consider a team trying to improve delivery. One option is to track time spent on tasks. Another is to track tasks completed that meet a defined acceptance standard. Time spent is easy to collect and can reflect effort, but it does not tell you whether useful work shipped. A long hour can be blocked, misdirected, or simply slower. “Five tasks completed that meet the definition of done” can be read directly. You do not need to infer. The measure names the thing the team is trying to produce.

This does not make time irrelevant. It can be a supporting diagnostic when delivery slips. But as a primary measure, completed work that meets the bar lands closer to the goal and avoids the pressure to inflate hours or stay late for the number.

Quick checks for directness

When you consider a candidate measure, test it with a few simple questions:

• Can a stranger tell whether the number moving is good or bad without a backstory?
• Does the measure name the thing you want, not something adjacent to it?
• Could someone hit the number while missing the goal? If yes, it is a warning.
• Would the same reading mean the same thing next week, or does it depend on context you often lack?

If a measure passes most of these checks, it likely reflects the goal directly enough to be trusted as a primary signal. If it fails, keep it as a supporting diagnostic or drop it.

Choosing between two candidates

When you are stuck between two measures, pick the one that needs the fewest words to explain and that cannot be improved by gaming behavior unrelated to the goal. You can always keep the other as a tool for debugging, not as the headline you chase.