THE SEMANTIC SEO OBSERVATORY
We rebuilt our website around how search engines understand businesses. Now we’re publishing what happens next.
On August 29, 2026, WebDevAuto completed a sitewide semantic SEO redesign. Beginning August 30, we started measuring the effect across search visibility, qualified traffic, leads, and business outcomes.
Loading the latest report
- Redesign completed
- August 29, 2026
- Measurement began
- August 30, 2026
- Latest completed data
- Still processing
- Last successful refresh
- Not published yet
Google Search Console publishes each day two to three days late. This page reports the latest complete day it has published, not today, and it refreshes every six hours rather than continuously.
THE EXPERIMENT
What is being measured, and when
Four dates define this experiment. Everything on the page is anchored to them, and none of them move.
August 1 to 28
Frozen pre-redesign baseline
The last 28 complete days before the rebuild. Once each source has published through August 28 its numbers are stored and never recalculated, so the comparison cannot move under its own feet.
August 29
Sitewide semantic redesign completed
The rebuild shipped during this day, so it belongs to neither period and is excluded from both sides of every comparison on this page.
August 30
Post-redesign measurement began
The first day whose traffic saw only the rebuilt site. Everything reported as "now" starts here.
Pending
Latest complete data published
Search Console has not published a complete day for this property in this refresh yet.
THE SCORECARD
Search visibility, traffic quality, business outcomes, semantic footprint
Every measurement states its own source and the exact dates on both sides of the comparison. A value that is not available says so; it is never filled in with a zero.
The latest measurements load with this page
The figures come from a report rebuilt every six hours from Search Console, Google Analytics, our own first-party events, and the CRM. The experiment timeline, the record of what changed, and the methodology below are static and do not depend on it.
DAY BY DAY
The daily record, across the redesign
One metric at a time, with the redesign date marked. Days a provider has not finished reporting are left blank rather than drawn as zero, and the line breaks across them rather than being interpolated.
No complete days to chart yet
The chart is published once at least one complete day exists at a provider. Until then there is nothing to draw that would not be invented.
TOPICAL AUTHORITY
Is Google understanding more of the business?
Search Console performance grouped by the site's own topic clusters. The seven capability clusters are the same records the navigation and every product page read from, so these are the site's real structure and not categories invented for a chart.
Semantic SEO is not only about ranking one page for one keyword. It is about helping search engines understand the full relationship between a business, its services, the problems it solves, and the markets it serves.
Not enough data yet. Cluster movement is published once Search Console has completed a post-redesign period to compare with the frozen baseline.
PAGE LEVEL
Which pages moved
A page has to clear a published minimum-impression threshold before it can be called a gainer, so a move from one impression to two can never outrank a real gain.
Gaining visibility
Pages with a verified increase that clears the minimum-volume floor.
No page has cleared the minimum-volume threshold with an increase yet. That is a statement about the amount of data, not about the pages.
Still developing
Pages Google is showing, without enough evidence yet to say they improved.
Nothing to list yet. Pages appear here once they are receiving impressions.
THE CHANGE
What we changed
Only work that is verifiable in the repository and in the live site is listed here.
The site was reorganised so that each topic has one page that owns it, and so that the relationships between those topics are stated in the structure rather than left implied. Thirty URLs moved, each with a single permanent redirect to the same topic; no page was deleted and nothing was rolled up into a hub.
The seven capabilities became root-level cluster heads with their supporting pages beneath them. Comparison content was consolidated under one path and reference content under another, so a single search intent is answered by a single URL. Where two pages were competing for the same question, one was made the owner and the other now redirects to it.
A single catalogue record per capability is now read by the header, the footer, the home page, and every page that describes the product, which is what makes the internal linking contextual rather than navigational: each capability page states how it connects to the other six and links to them on that basis. Headings, metadata, and breadcrumbs were brought into line with that structure, and every page in the sitemap is prerendered so a crawler receives the finished markup rather than an empty shell.
| Before | After |
|---|---|
| Capability pages sat under a /platform/ prefix that implied a hierarchy the content did not have. | The seven capabilities are root-level cluster heads, each the canonical page for its topic, with single-hop redirects from every old URL. |
| The product was described separately on the home page, the navigation, and each capability page. | One catalogue record per capability is read by the header, the footer, the home page, and every page that names it, so the site cannot describe itself two ways. |
| Internal links were mostly navigation: a page linked to whatever was in the menu. | Each capability page states how it connects to the others and links to them, so the links follow the relationships rather than the menu. |
| Comparison and guide content was split across /vs/, /compare/, and /guides/. | Decision content lives under /comparisons and reference content under /resources, one intent per URL. |
| Two URLs competed for the "website versus platform" question. | One page owns that intent and the duplicate 301s to it, so the two are no longer bidding against each other. |
| Performance was read as isolated keyword rankings. | Performance is measured by topic cluster, by page, and by what reached the CRM, which is what this page publishes. |
METHODOLOGY
How this experiment is measured
The frozen baseline is August 1 to August 28, 2026: the last 28 complete days before the redesign. The first time a source has finished reporting all 28 of them, those numbers are stored and never recalculated. A source that is not complete yet is recorded as incomplete with the reason, and no zero is substituted for it.
August 29 is excluded from both periods. The rebuild shipped during that day, so its traffic saw a mixture of the old site and the new one.
Post-redesign measurement began August 30, the first day whose traffic saw only the rebuilt site.
Matched-period comparison. Until 28 complete post-redesign days exist, the page compares the first N complete post-redesign days against the N days immediately before the redesign, where N is the number of complete days available from the slowest required source. Both sides always carry the same number of days.
Rolling 28-day comparison. Once 28 complete post-redesign days exist, the page compares the latest 28 of them against the frozen August 1 to 28 baseline.
Where each number comes from. Impressions, clicks, average position, click-through rate, query counts, topic clusters, and page movement come from Google Search Console. Organic sessions, engaged sessions, engagement rate, average engagement time, and landing-page counts come from Google Analytics, filtered to the Organic Search channel. Page and session events come from our own first-party analytics, which uses no cookies and stores no cross-site identifier; it is site-wide rather than organic-only, so it is reported for source health and never stands in for an organic figure. Lead, qualification, and booking outcomes come from the CRM.
Search Console reports late. A given day is typically published two to three days afterwards, and the lag is not constant. Nothing here assumes a fixed delay: every window is cut to the newest date the provider actually returned.
Correlation is not cause. A metric moving after a change is an observation. Naming the change as its cause requires ruling out everything else that moved at the same time, and this experiment does not do that.
Qualified leads matter more than traffic. Impressions are the cheapest number on this page to move and the least valuable. A qualified lead is the one that pays for the work, which is why it is reported even when the answer is that nothing changed.
Exclusions. Google Analytics figures use the property's own channel grouping, which is what separates organic search from paid, social, and referral traffic. Our first-party analytics does not record known bots, headless browsers, or our own prerender pass. This page's own URL is removed from every Search Console figure above, so a page that reports on the experiment cannot inflate it. Because those figures are summed per page rather than taken from the property total, they sit slightly above the totals shown inside Search Console itself. One search result can surface more than one page from this site, and the property total counts that as a single impression while the per-page breakdown counts one for each page. Across the frozen baseline the difference is about five percent.
Corrections. Providers restate historical data. Figures outside the frozen baseline may therefore change when a provider revises them, and the frozen baseline will not.
Branded and non-branded queries are separated by a stored list of brand spellings, matched with spacing and punctuation removed so "web dev auto" and "webdevauto" count as one brand. Queries that match no topic cluster are counted in an "Everything else" bucket rather than discarded, so the totals remain complete.
This page reports observed performance before and after the redesign. Search performance can also be affected by seasonality, algorithm updates, brand activity, competition, and demand. We do not label every change as being caused solely by the redesign.
DATA INTEGRITY
Where each number came from, and how current it is
If one source fails, the rest of the page still publishes and the failed one is marked. A failure never resets a previously published value to zero.
Source health is published with the first completed refresh.
Questions about this experiment
- Is this page live data?
- No, and it would be wrong to call it that. The report is rebuilt every six hours from each provider. Google Search Console itself publishes a given day two to three days late, so the newest figures here describe the latest complete day it has finished reporting, not today.
- Does the redesign cause the changes shown here?
- This page reports what was observed before and after a specific change on a specific date. Search performance is also affected by seasonality, algorithm updates, brand activity, competition, and demand, and none of those stop happening because we ran an experiment. We publish the observation and the windows it covers, and we do not label any single movement as caused by the redesign.
- Why is the baseline frozen?
- Providers restate recent days. If the baseline were recomputed on every refresh, a published percentage could move without anybody changing anything. So the last 28 complete days before the redesign are stored once, the first time a source has finished reporting all of them, and never recalculated.
- Why is August 29 missing from both periods?
- The rebuild shipped during that day, so its traffic saw a mixture of the old site and the new one. Counting it would attribute a part-old, part-new day to one side of the comparison, so it belongs to neither.
- Why compare four days against four days instead of against the whole month?
- Because comparing a partial period against a complete one is the most common way an SEO report flatters itself. Until 28 complete post-redesign days exist, the page compares the first N complete days against the N days immediately before the redesign. Both sides always carry the same number of days.
- Do you publish the search queries themselves?
- No. Very low-volume queries can contain personal detail somebody typed into Google. The page publishes counts, topic clusters, and page-level totals; it never itemises individual queries, and aggregate totals still include every row so nothing is quietly dropped.
- What happens if one of the data sources fails?
- The other sources still publish and the failed one is marked unavailable with a plain explanation. A failure never resets a previously published number to zero, and a refresh that cannot complete leaves the last verified report in place rather than replacing it with an empty one.
We don’t ask clients to trust an SEO theory we won’t test on ourselves.
WebDevAuto connects website structure, search visibility, lead capture, and customer outcomes in one system. This public experiment shows how we measure the work instead of hiding behind ranking screenshots.