Website analysis: map the pages, explain a variation

A site that no longer behaves as it used to raises two questions before any others: what is it made of, and what has moved? An analysis answers those two — the map of the site, then the reading of a variation.

What a website analysis brings under control

An SEO-driven analysis consists in measuring and explaining why a site gains, loses or stalls in organic visibility, then turning those findings into prioritized actions. It relies on observable data: a crawl for the structure, Search Console for impressions, clicks and indexing, audience data for engagement and conversions. The goal is not to pile up "warnings", but to identify the points that genuinely cost crawl, clicks or leads.

The three axes, and what each one gives you to observe

Three axes interact, and an analysis holds them together:

• Architecture and readability: URL structure, depth, hubs, internal linking logic, the site's ability to "show" its important pages to Google and to users.
• Content and intent match: alignment of page ↔ query ↔ intent, topical coverage, redundancy.
• Technical performance: the cost of loading and rendering the templates that carry traffic and conversion.

The first axis is the one the analysis handles in its own right: it presupposes a map of the site, and that is what is established here. The second is read at this stage as a gap — what a page targets against what it actually appears on. The third is measured, and that measurement — field data, root causes, the trade-off between cost of fixing and expected gain — belongs to the website performance audit.

Analysis and audit: a question of depth and rhythm

An analysis is an operational reading — site, pages, queries, results — used to decide quickly: what to fix, what to enrich, what to stop. An SEO audit formalizes the diagnosis and the roadmap further, and answers another question upstream: should you audit, and when. In other words, analysis serves to bring visibility under control day to day, while an audit structures a wider plan. The two complement each other, and rhythm makes the difference: a regular analysis stops you discovering too late a regression that would have cost several weeks of traffic.

Mapping the site: templates, roles and depth

An actionable map is not limited to "every URL": it groups by template and by role. It is the operation that makes everything else readable — without it, the figures cover averages that designate nobody, and the fixes land on whichever URL happens along.

The four page families, and what you attach to each one

Four families are enough to describe almost any site:

• Business pages: offers, strategic categories, demo or contact requests.
• Hubs: pillar pages that structure a theme.
• Supporting content: guides, FAQs, articles answering questions.
• Utility pages: legal, internal search, filters, pagination — to be controlled so as to avoid SEO noise.

Once the groups are formed, attach four values to each: its depth, its internal inbound links, its traffic and its visibility. It is that crossing which makes the difference between a list of URLs and a map: a map becomes actionable when it lets you decide "where" to fix — at template level — rather than "which URL" at random. A fix applied to a template propagates to all its pages; the same fix applied to an isolated URL is lost.

The hub, child pages, proof, conversion chain

The map above all serves to make one chain visible: hub → child pages → proof (cases, FAQ, comparisons) → conversion. Without that view, you often fix in the wrong place — optimizing an isolated page when the hub does not distribute authority, or producing new supporting content where the business page already lacks proof. Reading the chain means checking that each family plays its role: the hub directs, the child pages each cover one sub-intent, the proof answers an objection, the business page offers an action. A missing link shows up immediately on the map, while it stays invisible in a page-by-page report.

Depth and distribution of internal links

A common benchmark is to aim for around three clicks at most for the pages with something at stake. It is not a law, but a useful guardrail: beyond that, discovery slows down, internal linking weakens and pages become more "expensive" to crawl. Three questions are enough to read this part of the map: which pages receive the most internal links, and is that consistent with their business value? Which strategic pages are too deep, or connected through unintended paths — pagination, filters, internal search? Does internal linking make clear which page is the "reference" when several pages cover a close subject?

The same crossing brings out the orphan pages — those that no internal path reaches from the navigation or from contextual links, even when they appear in a sitemap or are known through campaigns. Compare the crawl's URL list with the URLs already known — sitemap, CMS exports, measured entry pages — and spot those that receive traffic with no internal link identified. The map names them; attaching them and deciding their fate belongs to the technical SEO audit, along with statuses, canonicals and the rest of the crawl signals.

The baseline: at constant scope, segment by segment

Nothing that follows holds without a point of comparison. A useful baseline compares at constant scope, and three cuts make up the essentials. The period first: period A against period B, same weekdays, same season, same campaigns — comparing a month with the previous one mixes seasonality with real effect. Brand queries against non-brand ones next: brand often hides structural falls, and a stable total can cover a net loss on everything that is not your name. Mobile against desktop last: the share of global web traffic coming from mobile is 60% (Webnyxt, 2026), and a degradation on mobile can be enough to sink conversions without any overall signal moving.

Two segments are added to these, less obvious and just as decisive. New against returning: acquisition content can look bad if you expect an immediate conversion from it. Strategic pages against long tail: do not drown your business pages in an average. Add the cut by area or by country if your business has one, and stick to it from one analysis to the next: a segment whose definition changes along the way makes the whole series unusable.

That leaves the discipline that makes the baseline replayable: always document the period and the segments included, and keep the exports. Without that, an "improvement" may be no more than a change of scope. And when the two periods are not comparable — a rebuilt template, an exceptional campaign, a modified tracking scope —, do not correct the figure: change the comparison. Isolate the subset that stayed stable, compare that alone, and rebuild a new baseline from the date of the change. An honest comparison on half the site is worth more than an overall variation nobody will be able to defend.

Explaining a variation: impressions, CTR, positions, conversions

A fall is rarely read with a single cause, and almost never at overall level. The approach consists in going down a notch — from the curve to the page, from the page to its queries — then coming back up to a family of cause, before mobilizing anyone at all.

Start from the page, not from the overall figure

The classic mistake is to analyse at overall level when SEO plays out page by page. Start from the strategic pages — offers, categories, hubs — then go back up to the associated queries, and compare three values together: impressions, clicks and CTR. A fall in clicks can come from a fall in CTR tied to the snippet displayed, not from a drop in positions. Then check positions on the queries with something at stake, not across the board: an aggregated average absorbs exactly what you are looking for. The click-through rate on page 2 of the SERPs is 0.78% (Ahrefs, 2025): the stake is therefore often to move queries from the second page into the top 10, and that is a reading done query by query.

The triangulated reading: three findings, three families of cause

To avoid false diagnoses, always cross two indicators rather than one. If Search Console shows a fall in impressions accompanied by a fall in positions, suspect relevance and intent, or an indexing problem — the trade-off between creating, optimizing or consolidating a page then belongs to the semantic audit. If impressions are stable but clicks are falling, suspect CTR: snippet, competition, AI Overviews, SERP format. If clicks are stable but conversions are falling, suspect the journey, the performance, the offer — or the tracking itself, which breaks more often than people think.

Inventory the changes, isolate seasonality

Before concluding, two checks avoid most attribution errors. First isolate seasonality and campaigns, by going back to the baseline: a variation that is found on the same date the previous year is not a problem to fix. Then inventory the on-site changes made during the period: templates, JavaScript, tracking, internal linking, and above all an unintentional noindex, which is the most mundane cause of the most spectacular falls. The table below sums up what each finding allows you to conclude, and what it does not.

Readability and journey: what the user understands, what they abandon

A well-ranked page can still produce nothing, and the analysis has to be able to record that without resorting to an aesthetic judgement. What is measured here is the gap between what the page promises in the SERP and what it delivers once opened.

Scannability: what a well-placed page misses

A page that is immediately understandable states its promise, structures it, proves it, then offers an action. In practice: informative headings and a hierarchy that reflects the real questions, menus that reduce the effort of choosing, and scannable sections — lists, boxes, tables — for readers in a hurry as much as for rich SERP formats. The stake is measurable: the share of clicks taken by the top 3 organic results is 75% (SEO.com, 2026), and the SEO statistics bring these benchmarks together with their source. A page that fails to be scanned in the first few seconds wastes that potential, even when it is well ranked.

Friction, and alignment between intent and action

Three kinds of friction show up in the data rather than to the eye, provided you do not confuse what is measured with what is inferred. Pogo-sticking — the visitor goes straight back to the results page, often because the page does not address the intent or hides the key information — appears neither in GA4 nor in Search Console: neither of the two sees a visitor going back to the SERP. It therefore remains a hypothesis, approached indirectly through weak engagement on visits coming from organic, a reduced visit depth and the comparison between templates serving the same intent. Weak entry pages, which read in both directions — highly visible but little clicked, the snippet does not convince; or heavily clicked but not engaging, the promise is not kept. And a confused journey: too many steps, contradictory actions, no reassurance, forms that are too long.

The useful reading then ties a page's intent to a measurable action: an informational page targets a micro-conversion, a comparison page quick access to the criteria for choosing, a conversion page an action with minimal friction. Spot the ones that attract visits but do not convert: that is usually a problem of intent, reassurance or journey, rarely of traffic. And to avoid bias from personal taste, steer by measured events and before/after comparisons on similar pages, on the same device.

The indicators that steer, and the ones that mislead

An analysis dashboard does not have to be complete: it has to be decidable. Five families are enough to cover what the site does; three very widespread metrics deserve, conversely, an explicit warning, because they set off more useless projects than any technical anomaly.

The minimum dashboard

Five families of indicators are enough to steer an analysis: visibility — impressions, share of queries in the top 10 and top 3, strategic pages exposed; CTR, by page and by query, where the gains can be significant, the impact of an optimized meta description on CTR being put at +43% (MyLittleBigWeb, 2026); indexing — the ratio between URLs submitted and indexed, abnormal exclusions, missing business pages; engagement — key events, scrolling, clicks on calls to action, by device; and conversions — requests, sign-ups, downloads, appointments, depending on your model. Each is read by group of templates, never as a site total, and each attaches to a page family on the map: that is what makes it possible to say which part of the site is progressing, and which one is absorbing the fall.

What the metrics do not prove

Three readings come up constantly and mislead just as often. A low performance score does not automatically imply a major SEO problem: it does not always explain an organic underperformance, and becomes a priority only if it affects strategic entry pages or if it holds back rendering and indexing. An aggregated average position hides the reality: steer by business pages and queries with something at stake, never by the site average. A rise in traffic can be unqualified, carried by non-strategic queries — in B2B, the quality of entries counts more than raw volume. None of these three reservations forbids acting; each one forbids concluding too quickly.

Governing the decisions that come out of the analysis

Without governance, an analysis produces contradictory optimizations — experience against SEO, content against product — and effects that are impossible to attribute. The minimum viable set-up comes down to three elements: a single backlog, dated and versioned, gathering every decision that comes out of the analysis; measurable acceptance criteria, based on the same sources as the diagnosis, with an observation period fixed in advance; and a written list of non-priority actions, which avoids scattering and the return of the same subject every quarter.

Each decision is documented in the same format, and that format is short: the pages or templates concerned, the evidence that triggered it, the acceptance criterion expected, and the before/after measurement plan. The requirement is not bureaucratic: with no evidence attached, a decision is re-debated at every meeting; with no observation period, it is judged too early; with no measurement plan, it is never judged at all.

The critical step remains the trade-off, and it is pronounced in three parts: what you fix now, what you postpone, and what you will not do — with its justification. This last point is the one people skip, and it is the one that costs the most. The analysis stops where the bottleneck becomes structural: when the quick wins have been taken and what remains depends on the templates or on the organization, the question is no longer how to read the site but how to change the way it is produced. Obtaining that map grouped by template, with each group's depth, internal inbound links, traffic and visibility, is the task covered by the audit and mapping module.

FAQ: frequently asked questions about website analysis

What is an SEO-driven analysis and what is it for?

It is a measurement and diagnosis exercise that connects architecture, content, technical performance and results — impressions, clicks, conversions. It serves to explain variations in visibility and to designate the family of cause before committing resources. Its product is not a list of anomalies, but a map of the site and a small number of documented decisions.

Which steps should you follow for a complete analysis without an exhaustive technical audit?

Start with the map: group the URLs by template and by role, with their depth, their inbound links, their traffic and their visibility. Then establish a segmented baseline at constant scope. Go down from the page to its queries to explain the variation observed, then settle on five to ten decisions with a success criterion attached. Technical exhaustiveness is not needed in order to decide.

How do you produce a reliable, actionable map?

Group the URLs by template and by role — hubs, business pages, supporting content, utility pages — then attach to each group its depth, its internal inbound links, its traffic and its visibility. The map becomes actionable when it lets you decide "where" to fix, at template level, rather than "which URL" at random.

How do you run a UX and SEO assessment without bias from design preferences?

Define behavioural metrics — events, scrolling, clicks on calls to action, engagement rate —, segment by device, and compare before/after on comparable pages and periods. You are looking for measurable evidence, not an aesthetic opinion. A page people dislike but that converts is not an analysis problem.

How do you analyse SEO page by page and by search intent?

Start from a page, list its main queries, classify them by intent — information, comparison, transaction — then check whether the page answers in the format the SERP expects. Then cross with audience data to validate traffic quality and conversion. It is that reading, and not the site average, that designates the page to handle.

How do you find out your traffic and interpret its variations with the right segments?

Analyse the entry pages, and segment systematically: brand against non-brand, mobile against desktop, new against returning, strategic pages against long tail. Then explain the variation by crossing impressions, CTR and positions with the changes made to the site — templates, internal linking, tracking. Counting sessions says nothing about what moved.

How do you fix orphan pages without breaking the architecture?

Avoid adding links everywhere. Create links from topically close parent pages — hub, category, guide —, add contextualized sections, then check that the page regains a logical path from the home page. If the page has no value, prefer a clean-up decision: deindexing or clean removal.

Continue reading

• The analysis has designated a family of cause and you now need the order of operations: see how to perform an SEO audit, from framing to prioritization, with the time to allow for.
• You redo the same analysis every quarter and rediscover the same regressions: SEO tracking installs KPIs, annotations and the control of gains over time.
• The map shows that your business pages receive few links, and the question becomes that of the links themselves: internal linking covers hubs, depth and the distribution of authority.
• The fall affects indexing and you first have to check that your pages are discovered: submitting a website covers the discovery, crawl and indexing chain and the ways to secure it.