An SEO experiment log records what changed, why it changed, and what the evidence showed afterward. For a small business, it can prevent repeated guesses and forgotten decisions. It does not turn every before-and-after comparison into proof that an edit caused a ranking or lead increase.
Reconstructed archive: November 2025. Written in November 2026 using contemporary historical sources. The archive date organizes this series; this text is not a recovered original.
The November 2025 addition
Google announced custom annotations in Search Console performance reports on November 17, 2025. The notes added contextual dates directly to charts and were visible to people with property access. They provided a reference for investigation rather than proof that an event caused the next movement.
Current practical expansion: The following buyer and implementation guidance was added in November 2026. It extends the archive discussion and is not presented as recovered original content.
Separate a change record from a controlled experiment
Many small websites do not have enough comparable pages or traffic for a meaningful controlled SEO test. A dated record is still useful, but call it an observation when that is what the method supports. Do not attach scientific certainty to a single page edited during a busy season.
A controlled design needs a suitable comparison and a plan for analyzing the result. A simple business log instead preserves the question, implementation, observations, competing explanations, and next decision. Either approach benefits from clarity about its limits.
Keep routine repairs distinct from discretionary experiments. A broken inquiry form should be repaired because it is broken, not left faulty to produce a comparison group.
Write the question before changing the page
Begin with a customer or search problem you can describe concretely. “Make the site better” cannot guide a useful test. “Visitors may not understand that this service excludes repairs” creates a clearer investigation.
Record the proposed change and why it might address the issue. Identify the page, audience, and expected observable result. Keep business facts accurate in every version; experiments are not permission to exaggerate availability, credentials, or results.
Choose a primary measure before examining the outcome. If the purpose is better inquiry fit, a rise in total visits alone does not answer the question. Keep secondary measures for context rather than switching to whichever number improved.
Use these fields for each entry
- Question: the problem and evidence that prompted the work.
- Scope: exact pages, services, and visitors affected.
- Before state: saved copy or screenshot and relevant baseline.
- Change: what will be different and why.
- Owner and date: who approves, implements, and verifies it.
- Measurement: metric definition, source, filters, and comparison approach.
- Other changes: releases, campaigns, pricing, staffing, or seasonal factors.
- Review decision: keep, revise, restore, or gather more evidence.
Store records in a format the business controls and can hand to a future provider. Avoid putting private customer details in a general marketing log. Aggregate counts and non-sensitive observations are often enough for the decision.
Preserve the baseline and implementation evidence
Save the report settings, not just a cropped chart. Record the period, page filters, device or country filters where relevant, and the definition of the outcome. Keep preliminary data and incomplete periods out of conclusions that assume a finished comparison.
Google’s Search Console guidance describes monitoring search performance and site issues. Use the appropriate reports as one source of evidence alongside website behavior and operational lead records. These systems describe different stages and need not produce matching totals.
After deployment, check that the intended change actually reached the public page. Record the verified date, especially if it differs from the drafting or approval date. A test of unpublished copy cannot explain a change in customer behavior.
Choose a review period that fits the evidence
There is no universal number of days that makes every SEO comparison reliable. Consider traffic volume, the sales cycle, seasonality, and the time needed for search systems to revisit changed content. Agree on an initial review checkpoint without promising that it will produce a definitive answer.
If very few relevant inquiries arrive, show the counts and the uncertainty. Extending observation may help, but it does not remove unrelated changes or create a controlled experiment after the fact.
Google’s traffic-change guidance identifies multiple causes of movement, including technical issues, demand changes, and ranking changes. Keep those alternatives in mind when an outcome happens after your edit. Timing alone is not a causal explanation.
A worked example without an invented success story
Imagine a service page that attracts inquiries for repair work the business does not offer. The proposed change adds a clear installation-only scope statement near the inquiry route. This is a hypothetical planning example, not a Guardian client result.
The log would identify the page, preserve the previous wording, record the verified release date, and define how staff classify inquiries. The main observation could be the number and share of assessed inquiries requesting an unoffered service, with pending and unknown cases shown separately.
During the review, the owner would also note a simultaneous advertising change or new phone-handling process. If unsuitable inquiries declined, the report could say the change was followed by an improvement while acknowledging those competing explanations. It should not claim the sentence alone produced the entire result.
Make a decision even when the outcome is uncertain
Not every entry needs a winning result. You may retain accurate clarification because it helps customers, while concluding that there is insufficient evidence of a measurable performance effect. You may reverse a layout change that makes a contact route harder to use even before a long-term report is ready.
Record the reasoning. “Kept because the service facts are clearer; effect on qualification remains uncertain” is a useful conclusion. “Successful optimization” without evidence is not.
If the team changes the hypothesis or measurement method, preserve the earlier entry and explain the new one. Do not silently rewrite the original goal to fit the outcome.
Use the log in the monthly SEO review
Review completed changes, unresolved tests, and decisions that need the owner’s input. Ask what the team learned and which next action follows. The record should reduce repeated work and make handovers easier, not become another administrative artifact nobody uses.
When comparing ongoing SEO support, ask how changes and evidence are documented. Our proposal evaluation checklist helps clarify implementation and reporting responsibility.
Talk with Guardian Solutions about a practical way to connect your website changes, performance observations, and monthly SEO decisions.
Historical source retained: Google announcement dated November 17, 2025.