Overview / Reading results

How to read a result

A short guide to the interpretation notes we write, and to reading any data summary with a healthy amount of caution.

Question 1

What does this figure cover?

Check the period, the population and the exclusions. A total of active customers that excludes accounts opened this month is a different number from one that includes them.

Question 2

Is the change bigger than the usual wobble?

Small categories move a lot from month to month. Our notes compare a change with the normal range seen in earlier periods before calling it a trend.

Question 3

What was set aside?

Exception counts matter. If 4.6% of records failed a rule, the result describes the other 95.4%, and the note says whether the missing part is likely to differ.

Question 4

What can it not tell us?

Correlation in a table is not a cause. Data recorded for one purpose may be blind to another. The last section of every note is about this.

Anatomy

The parts of an interpretation note

Headline in one sentence

The single most important thing, stated without adjectives.

What moved

The three or four changes large enough to discuss, each with its size and comparison period.

Coverage and gaps

Rows included, rows set aside and why, and any source that arrived late or incomplete.

Assumptions

Any rule that affects the reading, such as how refunds or part-months were treated.

Open questions

Things worth checking with the people who run the process, because the data alone cannot settle them.

Words we avoid

We avoid "proves", "guarantees" and "always". We prefer "is consistent with", "suggests" and "we cannot tell from these records". This is not hedging for its own sake: it keeps the note accurate and helps the reader judge how much weight to put on each figure.

Interpretation notes are descriptive. They are not financial, legal, tax or investment advice.