How to Analyse SEO: Evidence, Reporting and Action Plans

SEO planning distinguishes analysis from implementation. Analysis investigates the current condition; optimisation changes something in response. Define the question first: are important service pages missing, did clicks decline after a redesign, or are visits failing to become suitable enquiries? These questions require different evidence. Identify sites, languages and page groups in scope, plus areas excluded from review.

Record collection dates, comparison periods and relevant releases. Seasonality or a measurement change can affect a before-and-after comparison. Obtain authorised Search Console and analytics access through an appropriate process; do not put credentials or customer records in a brief. Visibility, an accepted enquiry and a sale are different stages. Establish which business outcome you can measure and document gaps.

Technical SEO auditing compares status codes, robots.txt, response directives, noindex, canonical, sitemap and internal links on important URLs. An intentionally excluded private dashboard and an accidentally excluded service page are different findings. JavaScript can change the rendered content, so compare appropriate output rather than assume a good browser appearance proves search access. Record crawler configuration, limits and samples. A partial crawl cannot establish that every page is healthy.

Google’s URL Inspection guidance distinguishes indexed information from the live test. Live access does not guarantee indexing or ranking. Unexpected canonical selection, blocked resources and content suitability require different investigation. Map redirects to relevant destinations rather than redirect every missing URL to the homepage. Verify a correction on representative pages and exceptions after release. Submitting a sitemap does not independently repair incorrect output.

Content analysis asks what the query means and whether the current page answers it. A service offer, product category and comparison guide may support different intentions. Titles, H1 and section headings should describe the page clearly. Do not force a universal keyword-density target or add supposed LSI keywords mechanically. Character counts are useful editorial checks, rather than a reason to ignore a clear and accurate description.

Build an inventory that records the information owner, update source, genuine examples, image descriptions and relevant links. Two pages appearing for a query are not automatically harmful cannibalisation. Compare their tasks and evidence before deciding to retain, improve or consolidate them. Link categories, services and guides where the next step helps the reader. For language equivalents, plan international SEO around consistent operational facts. More text cannot by itself establish complete coverage or expert evidence. Identify missing information and the person qualified to supply it before requesting another draft.

Audit scope should drive tool selection. Search Console provides your search information; GA4 provides configured visitor and event data. They are not interchangeable measures. PageSpeed Insights and Lighthouse support performance investigation, while Screaming Frog supports URL crawling. The current SEO Spider page describes a 500-URL free limit and restricted advanced capabilities. Crawling that allowance does not prove the whole site was reviewed.

The original guide also named Ahrefs, Semrush and Moz research tools, plus Surfer and Clearscope content assistants. Evaluate them for an actual requirement rather than a supposed universal winner. Similarweb can inform competitor traffic hypotheses; estimates are not access to a competitor’s analytics. Confirm current plans, data scope, seats and export needs before purchase. Vendor health and authority scores are not a shared Google scale. A licence is useful only if the team can interpret its evidence and implement a suitable response. Avoid turning a subscription list into the deliverable itself.

Performance review should test navigation, forms, product selection and reading on representative screens. A green laboratory score is not the entire user experience. The Core Web Vitals guidance uses LCP, INP and CLS for loading, interaction and visual stability. Good targets are at most 2.5 seconds, 200 milliseconds and 0.1 respectively, assessed at the 75th percentile with mobile and desktop segmentation.

Field data reflects actual visits, while laboratory tests use controlled conditions. Where field data is missing, do not describe a laboratory run as every visitor’s experience. The retired Mobile-Friendly Test is not a mandatory current audit tool. Inspect image dimensions, font loading, third-party code and interaction delays across relevant templates. Re-test important actions such as forms and checkout after changes. Faster output cannot independently fix an unclear offer or poor lead quality. Record the device, page and conditions behind a finding so another reviewer can reproduce it without assuming every visitor has the same network or interaction.

Competitor research should distinguish business competitors from the pages appearing for a query. A marketplace, forum or independent guide may occupy the results. Evaluate missing topics against buyer needs and information you can provide accurately. Do not copy a competitor’s page or link list. Their call to action and form may suggest an idea, but do not reveal their actual conversion rate or customer quality. Keep observed behaviour separate from interpretation.

Record backlink source, context and paid relationships separately. DR and spam scores are not Google penalties. Google’s disavow guidance explains that most sites do not need the tool and incorrect use can harm performance. Consider substantial artificial links together with a manual action or its likelihood. Do not automatically submit files for every unfavourable vendor score. Investigate manual actions, technical changes, demand and measurement before attributing a decline to a single cause. The next action should address a supported issue rather than imitate another company’s acquisition pattern.

Scope and budget review needs an executive explanation of the priority issue, affected pages and next deliverable. Engineering needs reproduction steps and expected output. Business owners need implementation capacity and a suitable outcome definition. Technical work is not automatically high priority while content is medium. Consider evidence, affected scope, risk and dependencies together. A small editorial correction can matter more than an irrelevant crawler warning.

The table is an illustrative report structure, not a verified Prix case. Report impressions, clicks, qualified enquiries and won customers as separate measures. Forecasts and ROI calculations need defined costs and assumptions. Do not turn an assumed first-place ranking into certain revenue. Attribution alone does not establish causation. Preserve report dates, completed work, unresolved issues and measurement limitations. A dashboard screenshot is not proof that a recommendation was implemented.

Measurement implementation may be a separate dependency where data is incomplete. Bring access, content, engineering and reporting owners into the action plan. A small correction, template change and structural migration need different checks and rollback arrangements. Limit the first work list to available capacity. Do not promise that each correction produces ranking growth within thirty days. A delivery milestone and a search outcome are different commitments.

Set review frequency around release rhythm and risk. Major migrations, unexplained data changes and new templates may need extra checks. There is no universal six-month interval that suits every site. This article does not pretend to supply a separate Excel file or a fifty-item downloadable resource that has not been prepared. In your brief, share the site, problem, period, recent changes and implementation capacity. Agree the work after evidence has been assessed, then check the delivered output. Clear responsibility helps prevent an analysis from becoming a report nobody can act on.