The partial month trap: 19 days and 12 days sitting beside 31

If a 92 day window opens on the 13th of one month and closes on the 12th of another, two of its four months are incomplete. In the clawlaw.in export on our proof page, taken 2026-10-14 across 2026-07-13 to 2026-10-12, July holds 19 days of data, August holds 31, September holds 30 and October holds 12. Chart those four months as totals and July and October collapse for a reason that has nothing to do with the site. Print the day count beside each month and the problem disappears. Most published monthly charts do not print it.

Where the distortion comes from

A monthly total is a day count multiplied by a daily rate, and a chart of totals silently compares both at once. When every month in the window is complete, the day counts are close enough that the comparison is about the rate, which is what the reader thinks they are looking at. The moment one month holds 12 days and the one before it holds 30, the chart is mostly about the calendar.

It is worst at the two ends, which is also where the eye goes. The first month of a window is the baseline a reader judges everything against, and the last is the one they treat as the current state. Understating both is the specific failure a day count column prevents.

The same four months, read two ways

Month	Days with data	Clicks a day	Complete month
July 2026	19	59	No, window opens 2026-07-13
August 2026	31	154	Yes
September 2026	30	282	Yes
October 2026	12	349	No, window closes 2026-10-12

Read as clicks a day, the four months are directly comparable and the sequence rises at every boundary. Read as totals, October would be the worst looking month in the window while being the strongest. One column decides which of those two pictures a reader gets.

Why windows end mid month in the first place

Because exports are taken on the day somebody needs them, and because Search Console holds back the most recent days while they are still being revised. An export taken on 2026-10-14 does not have complete data up to the 14th, which is why this window closes on the 12th. So a partial final month is not sloppiness. It is the normal consequence of taking a reading at a moment that is not the end of a calendar month, which is almost every moment.

The front edge has a different cause. A 92 day window is a 92 day window; it starts wherever counting back from the end date puts it. Rounding the window to whole months instead would throw away real data to make the chart tidier, which is the wrong trade.

It is worth noticing which direction each of those forces pushes. The held back days at the end make the final month short, and the fixed window length makes the first month short. Neither is chosen by whoever draws the chart, and both work against the people reading it, because the two months a reader leans on hardest are the two the calendar has quietly shrunk. That is unusual: most reporting problems come from somebody making a choice, and this one comes from nobody making a choice at all.

Four rules that make a monthly chart honest

• Print the day count for every month, complete ones included, so the short months do not look like exceptions being explained away.
• Use a daily rate rather than a total whenever any month in the window is partial.
• Put the window dates and the export date above the chart, because the partial months are a consequence of both.
• Never compare a part month to the same part of last month without saying so. A 12 day October against a full September is a comparison of calendars.

For the wider question of which instrument to read in the first place, and what each of them is a record of, the comparison between Search Console and an AI visibility tracker is worth reading before a reporting pack is designed, because the two count different surfaces and only one of them has months in it at all.

How to spot the trap in somebody else's chart

Look at the first and last bars. If the chart covers a window whose dates are stated and the first and last bars are the shortest, check whether those two months are complete before reading anything into the shape. If the dates are not stated at all, the bars cannot be interpreted and the right next question is which window this covers and when it was exported. We sell measurement and the page writing that follows it, so we ask buyers to put that question to us as well, which is why both dates sit above our own table rather than in a footnote.

Questions about partial months in a Search Console chart

Why does my current month always look weak?

Because it is incomplete, and usually missing its most recent days as well while they are still being revised. Compare it as a daily rate, or wait for the month to close before putting it beside the others as a total.

Should I just use complete months only?

For a published claim that is the cleanest option, and it costs you the most recent weeks, which are often the most interesting. The compromise that keeps both is a daily rate with the day count printed.

Does the same problem affect weekly charts?

Yes, and it is harder to see. A part week at either end of a window behaves exactly like a part month, with the added complication that a week missing a Monday is missing a different amount of demand from one missing a Sunday.

How many days of data does Search Console hold back?

Enough that the last day or two of any export is not final, which is why the export date matters as much as the window. The practical rule is to end the window a couple of days before the export rather than on the export date.

Is a daily average not just a smaller number?

It is a rate rather than a quantity, and a rate is the only thing that can be compared across months of different lengths. The totals belong in the same table if the reader wants them, with the day count that produced each one.

One column, and the reason it is almost always missing

A day count column is trivial to add and it makes a chart slightly less impressive, which between them explain its absence. The first month gets taller, the last month stops looking like a collapse, and the line flattens a little at both ends. Everything the column does is in the direction of a duller picture, and all of it is in the direction of a picture a reader can check. Adding it to a report you have already sent is also a cheap way to find out whether anybody read the first version, because the shape will change and somebody will ask why.