The measurement you should stop doing to make room for better

4 days ago
3 min read

Updated: 2 days ago

Introduction

Measurement accumulates. A number gets added because somebody asked for it once, it appears in the report every month afterwards, and it is never removed because removing things requires a decision and adding them does not. Two years later the report has nineteen figures, three of which anyone looks at.

The cost is not the effort of collecting them, although that is real. It is that the important numbers are harder to find, the review takes longer and therefore happens less often, and nobody can say what the report is for. Subtraction is the cheapest available improvement to most measurement.

It requires going through the list and asking one question of each item.

1. The measurement you should stop doing fails the decision test

The question to ask.

What would you do differently if this number doubled or halved. If the honest answer is nothing, it does not belong in the report. Ask it of every line. Most reports lose half their contents to this question alone.

2. Anything nobody has mentioned in a year

The neglect signal.

If a figure has never been referred to in a meeting, it is not informing anything. Silence is evidence. Remove it and see whether anybody notices. Keep collecting it quietly for a quarter if you are nervous.

3. Metrics you cannot influence

The agency test.

Numbers describing things outside your control belong in context, not in a performance report. They generate discussion and no action. Keep them separately if at all. Market conditions belong in the commentary, not the table.

4. Duplicated measures

The redundancy.

Three variations on enquiry volume tell you one thing three times and make the page longer. Keep the one with the clearest definition. The others can be derived if ever needed. Derivation on request is cheaper than reporting by default.

5. Anything requiring more than ten minutes to produce

The cost test.

A number that takes an afternoon to assemble will be skipped in a busy month, which makes the series unreliable anyway. Either automate it or drop it. Half-collected data is worse than none. A gap in the middle of a series undermines every comparison across it.

6. Numbers whose definition nobody can state

The comprehension test.

If the person reporting it cannot say precisely what it counts, nobody can interpret it. Define it or remove it. Both outcomes are improvements. Write the definition in the meeting if you can.

7. Metrics inherited from an old tool

The legacy.

Reports often contain figures that made sense in software abandoned two years ago. They persist by copy and paste. Check where each number originally came from. Anything with no current source is already gone.

8. Anything measured at the wrong frequency

The cadence mismatch.

A quarterly measure reported monthly is noise dressed as information, and it invites reaction to nothing. Change the frequency rather than the metric. Frequently that is the whole fix.

9. Remove them one at a time

The method.

Take one out per review and see whether it is missed, rather than redesigning the report in one go. Anything missed comes back immediately. Nothing else does.

Be careful about removing a measure because it has been stable for a long time. A number that has not moved in two years may be a control rather than a candidate for deletion, and dropping it means you will not notice when it finally does move.

Conclusion

Go through your report and ask what you would do differently if each number halved.

Remove anything nobody has mentioned in a year, anything outside your control, duplicated versions of the same measure, figures that take an afternoon to assemble, numbers whose definition nobody can state, and metrics inherited from software you no longer use. Fix the frequency where that is the real problem, and take things out one at a time so anything genuinely missed returns quickly.