How To Read Data Archives

The three levels of an archive

Every dated archive worth using is built in layers, widest first. This one has three, and knowing which layer answers your question is most of the skill.

1. Year. A year page shows all twelve months at once as a single grid. Use it when you already know the date and want the value in one look.
2. Month. A month page lists each date as its own row with the weekday alongside. Use it when you want context around a date, or when the weekday matters.
3. Date. The individual entry, reachable from either of the levels above. This is what you cite.

Work inward, never outward. Starting from a search box and hoping to land on the right entry is slower than three deliberate clicks, and far more likely to land you on the wrong year.

Reading a year grid

A year grid puts days down the side and months across the top. Rows are numbered 1 to 31; columns run January to December.

To find one date, locate the day number in the left-hand column, then read across to the month column. The cell where they meet is that date’s entry. To read a single month in date order, read straight down its column instead.

On a narrow screen the grid scrolls sideways while the date column stays fixed, so the row you are reading never loses its label. That fixed column is what makes a thirteen-column table usable on a phone at all.

Reading a month table

Month pages list the newest date first, which is the opposite of the grid’s top-to-bottom order. That is deliberate: someone opening the current month almost always wants the most recent entry, and someone reading a closed month is usually scanning rather than starting at the first of the month.

Each row carries the date, the weekday and the value, plus its position in the month. The position column is there so you can tell at a glance whether you are looking at an entry from the start or the end of a period.

What an empty cell means

An empty cell means no record exists for that date. It never means zero, and it is never a formatting artefact. There are exactly three reasons for one here:

• The date does not exist in that month — 31 April, 30 February.
• The date falls outside the archive window, which opens on 1 Jan 2023.
• The sources for that date could not be reconciled, so nothing was published.

The third case is the one that matters. An archive that fills gaps to look complete is worse than one that shows them, because a filled gap is indistinguishable from a verified entry once it is on the page. If you hold a verifiable source for a missing date, send it to the desk.

Checking an entry against another source

Before relying on any entry, check it. The method takes under a minute:

1. Note the exact date in ISO form, year first, so there is no day/month ambiguity.
2. Find the same date in an independent source — not one that republishes this archive.
3. Compare the dates first, then the values.

That order matters. Two sources disagreeing about a value is a genuine conflict worth reporting. Two sources that appear to disagree but are using different date conventions are simply misaligned, and reading the second comparison first is how people convince themselves of a conflict that does not exist.

Four mistakes that cost people time

• Reading the wrong year. Grids look identical between years. The year is in the heading and in the URL; check one of them.
• Dropping a leading zero. 07 is a two-digit value. Stored or copied as 7 it stops matching anything.
• Treating a gap as a zero. A missing entry contributes nothing to a count. Counting it as 00 corrupts every statistic built on top of it.
• Reading counts as forecasts. Frequency describes what has been recorded. It says nothing about any future entry — see understanding previous records.