A monthly daily average cannot be chosen, and a good week can

Publish a monthly daily average and you have published a figure nobody could have selected. Publish a screenshot of a good week and you have published a figure somebody chose, whether or not they meant to. That is the whole case for the first and against the second. The four figures published for clawlaw.in on our proof page are monthly daily averages: 59 clicks a day in July 2026, 154 in August, 282 in September and 349 in October. They come from a Google Search Console export taken 2026-10-14 over the window 2026-07-13 to 2026-10-12. The site was not new when that window opened; it already carried pages and a search history, which is a fact that has to travel with any multiple read off these four numbers.

Why a month cannot be picked and a week can

A calendar month has a boundary somebody else set. There are only four of them in a 92 day window, so there are four candidate figures and a reader can see all of them at once. A seven day window inside the same 92 days can start on almost any day in it, so whoever picks one is picking from scores of candidates. The best of scores of candidates is a very different statistic from the average of all of them, even when nobody is being dishonest. A month boundary removes the choice. That is its entire value as a unit of evidence.

The second property matters as much. A daily average over a month absorbs the weekly shape of the business. Most sites have a weekday pattern, so a week that happens to contain a public holiday, a campaign or a news event moves a weekly figure and barely touches a monthly one. The monthly figure is therefore comparable to the month beside it in a way two weeks drawn from different parts of a quarter are not.

The four figures, and the column most tables leave off

Month	Clicks a day	Impressions a day	Days with data
July 2026	59	2,253	19
August 2026	154	7,276	31
September 2026	282	14,703	30
October 2026	349	18,331	12

The last column is the one that makes the table honest. July holds 19 days because the window opens on 13 July, and October holds 12 because the export was taken on the 14th and the window closes on the 12th. Using a daily average rather than a monthly total is what makes those two partial months comparable to the two complete ones at all. Had the same data gone out as four monthly totals, the two short months would have read as the weakest in the set, and the reader would have been looking at a calendar rather than at a result.

What a daily average still will not tell a reader

It is an average, so it hides the spread inside the month. It carries no split between brand searches and the rest, which matters because brand clicks follow your advertising rather than your search work. And it says nothing about which queries or pages produced the clicks. Each of those is a separate column a report can add, and none of them is a reason to go back to a weekly screenshot.

How to build the table in ten minutes

• Export daily rows from the Performance report for the whole window rather than reading the chart.
• Group by calendar month and divide each month's total by its number of days with data, not by the length of the month.
• Print the day count beside every row, including the complete ones, so the partial ones do not look like a special case.
• Put the window and the export date above the table, because a monthly average with no window is a figure with no scope.
• Keep clicks and impressions in separate columns. They are different units and sharing an axis makes the smaller one unreadable.

For what belongs in the scorecard above a table like this, and what should stay on the working dashboard underneath it, the SEO KPI scorecard splits the two documents in the way boards tend to accept.

Why we report it this way on our own work

Because we are asking buyers to apply a test to suppliers and we sell one of the services being tested, so our own figures go out in the form we say is the honest one. Two engines have said the same thing in their own words: asked on 15 October 2026 what would make a page worth citing, 101 of 102 conversations on ChatGPT and Gemini named a stated measurement window and 80 named the sample size. A monthly table with a day count column carries both without a sentence of explanation.

Questions about reporting growth as a monthly daily average

Does averaging over a month throw away the detail?

It sets the detail aside, which is the feature. What a monthly figure answers is whether the level of the business moved, and ordinary day to day swing is exactly what stops a daily figure answering that.

What about the first and last months of a window?

Print their day count and use the daily average rather than the total. A 19 day month and a 31 day month are directly comparable per day and not at all comparable as totals.

Should I show a rolling average instead?

A seven day rolling average is the right instrument for watching a line week by week while work is in progress. For a published claim the monthly figure is better, because a rolling window still has a start date somebody picked.

Does this work for a site with very few clicks?

At low volumes report the monthly total and the day count rather than a daily average, because an average below one per day invites a false precision. The principle is the same: the month boundary is not yours to move.

If a board will only look at one row, which one?

Non brand clicks a day by month, with the day count next to each. The business recognises the unit, nobody chose the periods, and it survives the question of where the numbers came from.

Four figures and one thing they cannot be

The sequence 59, 154, 282, 349 is a reading about one domain, in one window, in one market, and it is not a benchmark for anybody else. The multiple depends almost entirely on what the site already had, so the same four months of work on a different starting footprint would produce a different sequence and neither would be the wrong one. The reason to publish it is not to offer a target. It is to show what the unit looks like when it is chosen before the result is known, so that the next time somebody shows you a strong seven days you have a specific question ready about every other week they could have shown you instead.