What evidence makes an SEO renewal conversation easy?

An SEO renewal conversation is easy when both sides can check the evidence: a dated log of every change, comparison pages chosen before the results, each result with its size and interval, the inconclusive results, and a plain list of what stays uncertain. Anything less leaves each side arguing from memory.

• Google says no one can guarantee a #1 ranking and suggests asking an SEO what results to expect, in what timeframe and how success is measured (Google, Search Central, read 30 October 2026).
• Search Console keeps 16 months of data, 4 months more than a 12-month engagement lasts (Google, Analytics Help, read 30 October 2026).
• Search Console annotations are capped at 200 per property and 120 characters each, and are deleted after 500 days (Google, Search Console Help, read 30 October 2026).
• Penlid freezes up to 30 untouched pages as a change's comparison pool at the moment it is recorded, later narrowed to the best eight (Penlid, methodology, read 30 October 2026).
• A page with fewer than 10 clicks in the 28 days before a change gets a direction at best and never a size (Penlid, methodology, read 30 October 2026).

What can a client fairly ask an SEO to show at renewal?

A client can fairly ask what results to expect, in what timeframe, how success is measured and what was changed, because Google's advice on hiring an SEO lists those questions. They include What kind of results do you expect to see, and in what timeframe? and Will you share with me all the changes you make to my site. Google adds: No one can guarantee a #1 ranking on Google.

A renewal is when those questions come due. Our reasoning, not Google's: each answer needs a record made at the time, because a record rebuilt at the meeting is rebuilt knowing the outcome. A client with access to the property can also check figures personally, since Google's permissions table marks the Performance report as available to owners, full users and restricted users.

Which records make up a renewal pack?

A renewal pack holds seven records: a dated change log, fixed comparison pages, a result for each change, the inconclusive results, Search Console history, known data problems, and a list of what stays uncertain.

The Penlid methodology says the p-value, the interval and the number of untouched pages are printed under the verdict, and that the verdict never arrives without its working. A pack should hold itself to the same rule.

Why must the comparison pages be fixed before anyone reads the results?

The comparison pages must be fixed before anyone reads the results, because a set chosen afterwards can quietly decide the answer. Penlid freezes its pool when a change is recorded, and the methodology says why: the one fact that cannot be rebuilt afterwards is which pages you left alone.

The frozen pool holds up to 30 untouched pages from the same site, and the best eight are used. Which of them qualify is read from each page's own clicks on days −56 to −29, a stretch kept apart from the days the lift is measured over, except that the size check also reads the changed page's clicks from the 28 days before the change. A further exception is stated: when more than eight qualify, their order also uses a trend over the 56 days before the change, which includes days the lift is measured against.

Without Penlid's paid product, the same rule works by hand: write the list down, with the date, when the change goes live. The free did-it-work tool takes up to 50 changes and, on a day-by-day export, chooses the comparison pages itself. On identical counts it gives the same result as the paid product, so a client with the export and the change list can re-run a check. Its result can be copied as text or printed or saved as a PDF, and its page says those are the only two things that leave its memory. For choosing comparison pages by hand, see control pages.

How should an inconclusive result appear in the pack?

An inconclusive result should appear in the pack as a result, together with the smallest lift the measurement could have detected, because that figure shows what the test was able to see. Small pages produce many such results: in Penlid's simulations a page with 4 clicks a day, about 112 in 28 days, was given a known lift and measured against untouched pages.

A real change smaller than these lifts will often come back "we cannot tell". Below 10 clicks in the 28 days before the change the answer is a direction at best, never a size: Penlid tested a threshold of 5 clicks and kept 10, because intervals for 5 to 9 clicks covered the true value 97.78% of the time on average but only 89.65% in the worst scenario. Our reasoning: say this at the start of the engagement, so an inconclusive result is not heard as "nothing happened". See inconclusive SEO tests.

Why should the pack list every change that was measured?

The pack should list every measured change, not only the wins, because the more changes are tested the more likely it is that some look like wins by chance. Across 675 simulated scenarios with nothing done to the page, Penlid's estimator called a change significant 3.26% of the time on average, where a 95% test allows 5%; the worst single scenario was 5.85%.

The methodology calls refreshing pages because they were declining the largest remaining source of false "improved" results. Pages whose last 28 days sat at least one standard deviation below their own average, then left alone, came back "improved" 5.50% to 9.20% of the time, depending on drift between pages, against at most 3.10% for pages not picked that way. So the log should say why each page was chosen, and an "improved" on such a page reads as at least as good as the pages left alone, and possibly some recovery that was coming anyway. See regression to the mean and false wins.

Which history and known data problems belong in the pack?

The pack needs Search Console history for the whole engagement and any known Google data problem inside it. Search Console keeps 16 months, so at a 12-month renewal its window still reaches back 4 months before the engagement began (16 − 12 = 4), at 16 months it reaches back to the start and no further, and at 18 months the first 2 months of the engagement are gone (18 − 16 = 2). Annotations make a poor change log, since Google deletes them after 500 days, about 16.4 months (500 ÷ 30.4 = 16.4). In our reasoning, start saving the data and the log on day one; the guide to the 16-month limit lists the options.

Google's data anomalies page lists known reporting errors. As read on 30 October 2026, it says impressions, CTR and average position were not accurately reported from 13 May 2025 to 27 April 2026, and that clicks were not affected by that error. Our reading: a pack covering those dates should compare clicks and say so, and should list any other entry on that page that touches the pages compared. For a drop in the period, see what to check first when traffic drops.