Why the best day is not the last day on a new domain

A new index entry settles. Our own export records 2,685 impressions on 11 October 2026 and 2,055 on 12 October 2026, and the second figure is in the published series on purpose. A chart whose last bar is also its tallest tells you something about whoever picked the end date, and very little about the site. The aiknowsus.com window on our proof page runs 2026-10-04 to 2026-10-12 and the export was taken 2026-10-14, so the end date is the last day with data rather than the best one available.

What settling after a peak actually is

When a batch of pages becomes eligible across a wide set of queries at once, the first days at the new level are the engine working out where those pages belong. Some queries it served you for stop including you. Others start. The level finds a resting point somewhere near but usually below the first day of it. That is not decline and it is not decay. It is the difference between eligibility arriving and eligibility settling, and on a nine day old domain it happens inside days.

Reading one day down from a peak as a drop requires treating a single day as a direction, which no daily figure can support. A direction needs enough days that the ordinary movement between them is smaller than the change you are claiming.

How to tell a settle from a real fall

Reading	Settling	A genuine fall
Where the level lands	Near the peak, well above the previous level	Back at or below the previous level
Query count	Holds or keeps widening	Falls with the impressions
Position average	Moves either way as new queries settle	Worsens alongside the impression fall
How long to wait	Days. The new level is visible quickly	Diagnose it, because waiting does not fix a fall

On our own series the first column is what happened: 2,055 sits far above the 84 of two days earlier, and the previous level was in single digits. For a site with history rather than nine days of it, the diagnosis order for a real decline is a different job entirely, and how to measure organic traffic, read it and diagnose a drop runs through it in the order that finds the cause fastest.

The end date is a choice, and it is usually invisible

Every published chart has an end date and almost none of them say how it was picked. There are only three honest answers: it is the last day with data, it is the end of a calendar period agreed in advance, or it is the day the work finished. Anything else is a selection, and the signature of a selection is a final bar that is also the tallest one in the picture.

Three habits make the choice visible rather than suspicious:

• End on the last day with data and say so, which removes the decision entirely.
• Print the day count so a reader can see whether the final period is complete.
• Keep the dull days in. A series with no settle, no flat stretch and no step down reads as curated, and a careful reader notices that faster than they notice the headline.

Why this is worth a page on a nine day window

Because the temptation is strongest exactly here. A nine day series with a peak on day eight is the single most flattering chart this programme has ever produced, and stopping it one day earlier would have made the final bar the tallest and the line look vertical. The 2,055 costs us that picture and buys the only thing that makes the rest of the export worth quoting. It is also what separates a stated window from a chosen one, and that distinction survives however precisely the dates are printed: a reader cannot verify an end date, they can only notice when it coincides with the best available number.

What we would have to publish for the next reading to count

The same domain, a longer window, the same two columns, and the end date set by the data rather than by us. Anything else is a new chart rather than a second reading. Taking this measurement for other businesses and writing the pages behind it is what we are paid for, and a test we ask buyers to run on suppliers is worthless if our own figures have not been through it.

Questions about a fall after a peak in Search Console

My impressions dropped the day after my best day. Is that bad?

One day is not enough to read as anything. Set the new level against the level before the rise rather than against the peak, and check whether the query count held while it happened.

What length of window shows a genuine decline?

Whole weeks set against whole weeks, with the same number of weekends in each, and the query count alongside. A single day, a single weekend or a part week will produce whichever answer you went looking for.

Should I cut my chart at the peak if the client only wants good news?

Give them the full series with the peak marked. The peak is still in the picture and it is now believable, which is worth more than a steeper line that invites one question nobody in the room can answer.

Does a settle mean the work stopped working?

No. The level after a settle is the result. The peak was the arrival, and arrival is briefly higher than the level that follows it because the engine has not yet decided which queries you belong in.

Which end date should a published window use?

The last day with complete data, said out loud, with the export date beside it. Both of ours are printed, which is the only reason a reader can check that the end was not picked for its height.

The chart nobody can end early

One useful property of a series that covers the entire life of a domain is that it cannot be trimmed at the front, and once the end date is fixed as the last day with data it cannot be trimmed at the back either. That leaves no degrees of freedom at all, which is an unusual and strong position for a published figure to be in, and it is the reason a nine day export from a nine day old site is worth more to a sceptical reader than a longer window with two adjustable edges. When the next window is published the figures will be worse looking in some respects and the shape of the chart will be duller. Both of those are the point.