gtmjosh
Report on What Matters
Chapters
  1. 01 · Ask a Business Question, Not for a Dashboard
  2. 02 · Choose the Right Population and Denominator
  3. 03 · Separate Volume, Speed, Quality, Cost, and Outcome
  4. 04 · Compare Trends and Useful Segments
  5. 05 · Use Leading and Lagging Indicators
  6. 06 · Show Sample Size, Missing Data, and Blind Spots
  7. 07 · Connect Summary Metrics to the Records
  8. 08 · Design for the Person Using the Report
  9. 09 · Capstone: Build a Reporting Brief
Guide overview →

12 min · Interactive exercise

Chapter 7 of 90 complete

Connect Summary Metrics to the Records

Every summary opens, and one of them gets read on purpose.

Drill-through is a first-class requirement, not a nice-to-have. Every aggregate in the brief names how to get from the number to the records underneath it, filtered the same way.

The reason is not curiosity. It is that this is the only way anybody ever finds out a number is wrong.

The field that disagrees with the notes

This is the payoff for the entire previous guide and it is the most useful thing in this one.

In a real analysis, a categorical reason field reported that competitors accounted for a small share of losses. Somebody read the free-text notes on those same records and found a competitor named several times more often than the field claimed.

The field was not broken. It was filled in by people, at a moment when categorising a loss was the least interesting part of their day, and the default was easier than the accurate answer. A field is only as good as the process that fills it, which the previous guide spent eight chapters on, and the only way to discover the gap is to go and read the records.

So the brief should name, for at least one metric, the record-level check that would falsify it. Not as a one-off audit. As a standing part of how the number is trusted.

GTM Lab

Saved locally
Metric Dictionary and Reporting Brief · 0 of 9 sections started

Saved locally to your browser.

Open the number

The field and the records disagree, and the only reason anybody knows is that somebody read them. Add a path from every number to its records, then write the check.

Fixture: field-versus-notes

The field and the record disagree

FIXTURE-FIELD-VS-NOTES
What the reason field reports
A small share
What the free-text notes say
Several times more often
Was the field broken?
No
Who fills it in
People, at the least interesting moment of their day
How anybody found out
By reading the records
Resets every run.
What does drill-through actually buy you?
A reason field reports one rate and the notes on those records tell a different story. What is the conclusion?
Chapter 7 of 90 complete