What should a monthly SEO report contain so the client can trust it?

A monthly SEO report a client can trust defines each metric, shows its source and dates, and states how sure each result is. It lists known data gaps and what was changed and when, and it compares changed pages with pages left alone instead of printing a bare before-and-after number.

• A Search Console table shows at most 1,000 rows, and anonymized queries are left out of the rows but counted in chart totals unless a query filter is applied (Google, Search Console Help, read 30 October 2026).
• Google's data anomalies page lists impressions, CTR and average position as not accurately reported from 13 May 2025 to 27 Apr 2026, with clicks not affected (Google, Search Console Help, read 30 October 2026).
• Penlid's SEO significance calculator compares a changed page with three to eight untouched pages over two 28-day windows (Penlid, SEO significance calculator, read 30 October 2026).
• With 10 independent tests at the 5% level, the chance of at least one false positive is about 40%, from 1 − 0.95^10 (Wikipedia, Multiple comparisons problem, read 30 October 2026).

How should the report define each number?

Define every number in Google's terms and name its tool. Clicks are the times a user clicked through to your site from Google Search results, impressions are how often the site appeared in them, CTR is clicks divided by impressions, and average position is the topmost position a page held for a query, averaged across queries.

Position needs the most care. In Google's example, positions 2, 4 and 6 for one query count as 2, positions 3, 5 and 9 for another count as 3, and the average is (2 + 3) ÷ 2 = 2.5, a position the site never held. Google warns that position can mislead and recommends watching how it changes over time, so print it as a trend and label it an average, not a ranking.

Search Console covers how the site appears in Search up to the click, and Analytics covers what visitors do afterward, so clicks and sessions will not match exactly. Google says it's normal and expected that the numbers will be a little different. Print the gap each month so the client never finds it first.

Which dates and time zone should every number use?

Use complete days, Pacific Time as Search Console reports it, and periods of equal length. Weeks run Sunday to Saturday, months are calendar months, and recent periods can include preliminary data that Google says might change in the next few hours.

Calendar months are a trap, by our own arithmetic: a steady 30 clicks a day gives 840 clicks in a 28-day month and 930 in a 31-day month, so the longer one looks 10.7% better (930 ÷ 840 = 1.107) with nothing changed. Compare clicks per day or equal windows; Penlid's 28-day windows hold each weekday exactly four times.

Pull the data late and print the date: Google says collected data is usually available in 2 to 3 days. A change needs longer. Penlid's SEO test calculator reads a result 38 days after go-live (28 days of after window, a 7-day settling allowance that is ours, not Google's, and about 3 days of reporting lag).

Why do the rows in a table not add up to the chart total?

Table rows can add up to less than the chart total because a Search Console table shows at most 1,000 rows and leaves out anonymized queries, which the chart totals still count unless a query filter is applied.

Report the chart total as the headline, say how many rows each table lists, and print the gap. Google says the bulk data export gives the most complete list of queries; see the guide to the 1,000-row export limit.

Which entries on Google's data anomalies page should the report disclose?

A report should disclose every entry on Google's data anomalies page that overlaps its dates, because those entries can explain movements that no SEO work caused. As read on 30 October 2026, the page lists impressions, CTR and average position as not accurately reported from 13 May 2025 to 27 April 2026 (clicks were not affected, and the entry says the issue has been resolved), and FAQ rich results no longer appearing in Google Search from 7 May 2026 (reported FAQ impressions will drop).

The first entry sets a year-on-year trap, by our own arithmetic on Google's dates: impressions, CTR or position for any date from 13 May 2026 to 27 April 2027, compared with a year earlier, are set against a date inside the error window. Penlid's anomaly check notes that Google does not say whether an export downloaded today still carries the old figures, so compare clicks or warn the client. Re-read the page for every report, and see what to check when traffic drops.

How should the report record what was changed and when?

Record each change with its date, the page and the reason, in a log kept with the report. Google's advice to anyone hiring an SEO includes asking whether they will share all the changes they make to the site and the reasoning behind their recommendations.

Search Console annotations can mark a date on the chart, but they are not a record: a property holds up to 200, each note is up to 120 characters, they are deleted automatically after 500 days, and they do not show in comparison mode.

How should the report show whether a change worked?

Show each result as a lift against what pages left alone predict, with its 95% interval, p-value and number of comparison pages, and say plainly when the numbers cannot tell. Google says it can be hard to tell whether one change caused an improvement, because other events also move performance.

The SEO significance calculator takes the changed page's clicks for the 28 days before and the 28 days after, leaving out the day itself, and the same two figures for three to eight pages you left alone, chosen and written down before you look at their results. Each comparison page needs 10 or more clicks across the two windows. With a Search Console file, the did-it-work tool chooses comparison pages from your own data.

Say how the changed pages were picked. In the simulations on Penlid's methodology page, pages picked because they were declining and then left unchanged came back "improved" 5.50% to 9.20% of the time, against at most 3.10% otherwise, so an "improved" on such pages may include recovery that was coming anyway. See proving SEO results to a client and regression to the mean.

How many of a report's wins can be chance?

Some wins in a report can be chance, and more of them as it lists more changes. If none of n changes did anything and each is tested at the 5% level, the chance that at least one shows a false win is 1 − 0.95^n: 22.6% for 5 changes, 40.1% for 10 and 64.2% for 20 (approximate: pages on one site are not fully independent).

We suggest listing every change tested in the month, not only the winners, and re-checking a lone "improved" among many. See false wins in SEO testing.