Find what’s stopping your website from growing.

What we would investigate

1. 1

   Establish a clean baseline

   Pull the full available Search Console history and GA4, so seasonality and year-over-year comparison are possible before anything is concluded.

2. 2

   Find the exact date

   Day-level analysis per template and per query group. Different sections often have different dates, and that pattern alone eliminates whole categories of cause.

3. 3

   Check indexing and coverage

   Did pages leave the index, or did they stay indexed and lose position? These are completely different problems with completely different fixes.

4. 4

   Compare technical state before and after

   Canonicals, robots, status codes, redirects, rendering and internal links around the date. Most drops that are called algorithmic turn out to be a release.

5. 5

   Compare content changes

   Pages edited, consolidated, pruned or moved. Removing pages with real demand is a common self-inflicted cause.

6. 6

   Check whether the drop is real

   Cross-check GSC against GA4 and the CRM. A consent or tracking change can produce a drop in reported traffic while actual traffic is unchanged.

7. 7

   Compare competitors and result layout

   Did someone take those queries, or did the result page itself change shape so the same position now earns fewer clicks?

8. 8

   Check links to affected URLs

   Lost backlinks or internal links to the specific pages that dropped — measured on the affected cohort, not site-wide.

9. 9

   Line up external events

   Known search updates, seasonality, market events and your own campaign history against the isolated date.

What we would investigate

1. 1

   Check whether the leads exist

   Reconcile GA4 against the CRM first. A meaningful share of “we get no leads” turns out to be “we get leads and cannot see them”.

2. 2

   Classify the traffic by intent

   Split queries into brand, commercial, informational and irrelevant. Informational traffic converting at a low rate is normal, not a failure.

3. 3

   Measure conversion by landing page

   Conversion rate per template and per page, not site-wide. A site-wide average hides both the good pages and the broken ones.

4. 4

   Check the conversion path on each key template

   Is there a clear next step? Is it the right next step for that buyer stage? “Learn more” everywhere is a real cause of this problem.

5. 5

   Test the forms for real

   Submit them. Check validation, error handling, mobile behaviour, required fields, spam filtering and whether the submission actually reaches the CRM.

6. 6

   Analyse lead quality by source page

   If leads exist but sales calls them poor, the problem is which searches you are winning — not the conversion rate.

7. 7

   Compare against paid performance

   If paid converts on the same landing pages and organic does not, the difference is usually intent, not the page.

What we would investigate

1. 1

   Inventory everything that collects data

   Every GA4 property and stream, every GTM tag, trigger and variable, every ads pixel, and every CRM field that is supposed to capture source.

2. 2

   Classify every event and conversion

   Correct, wrong, duplicated or legacy. Most containers we open have tags nobody currently owns.

3. 3

   Check what conversions actually measure

   A conversion defined on a thank-you page view is not the same as a confirmed submission — and behaves completely differently when routing changes.

4. 4

   Perform real actions

   Submit a real form, complete a real purchase or signup, click a phone link, and trace each one through GTM, GA4, the ads platforms and the CRM.

5. 5

   Test both consent paths

   Granted and denied. A consent implementation that drops data instead of modelling it can move reported traffic overnight for a compliance reason nobody connected to the chart.

6. 6

   Check cross-domain and subdomain behaviour

   One journey becoming two sessions inflates traffic, destroys attribution and creates enormous “direct”.

7. 7

   Reconcile against the CRM

   Match front-end conversions against CRM records, quantify the gap, and explain each contributor to it rather than averaging over it.

What we would investigate

1. 1

   Build the URL inventory

   Crawl the old site, then add Search Console pages, GA4 landing pages, backlinked URLs and conversion pages. One list, every URL.

2. 2

   Score every URL by value

   Traffic, conversions, revenue, backlinks and query value. This is what turns a redirect map from a guess into a protection plan.

3. 3

   Map old to new as rules

   Rules wherever a pattern exists, explicit pairs where it does not. A rule can be tested against thousands of URLs; a spreadsheet cannot.

4. 4

   Specify what must carry across

   Metadata, canonicals, structured data, internal links, sitemaps and robots — for both the staging and production environments.

5. 5

   Specify analytics continuity

   Events, conversions, dataLayer and cross-domain behaviour on the new build, tested on staging before launch.

6. 6

   Test on staging

   Crawl staging, run the redirect rules against the real scored URL set, verify rendering and markup, validate the tags.

7. 7

   Define launch-day pass/fail criteria

   And rollback triggers. “Should we roll back?” should be answered by criteria agreed in advance, not by argument at 2am.

8. 8

   Monitor for 30 days

   Day 1 critical QA, day 3 indexing and errors, day 7 traffic comparison, day 14 page-level comparison, day 30 retained-versus-lost report.

What we would investigate

1. 1

   Segment the site by page type

   Not by URL. Every analysis after this — indexation, traffic, conversion, quality — is done per type, which is the only way large-site data becomes readable.

2. 2

   Measure indexation by type

   What share of each page type is indexed, and what share of the indexed pages earn anything within 90 days.

3. 3

   Map crawl budget

   Where crawl requests actually go, using server logs where available. Usually a large share goes to parameters, facets and duplicates.

4. 4

   Analyse the internal link graph

   Link depth and internal links received by the pages that carry commercial value, versus the pages that receive links because of the navigation template.

5. 5

   Find the duplication and cannibalisation clusters

   At scale these are systematic rather than incidental, which means they can be fixed with a rule instead of page by page.

6. 6

   Define a quality floor per page type

   The minimum data or content a page must have to deserve indexation. Then measure how many pages currently clear it.

7. 7

   Design indexation and canonical rules

   Which combinations are indexable, which are canonicalised, which are blocked, and how sitemaps are segmented so coverage is diagnosable.

What we would investigate

1. 1

   Build a fixed prompt set

   Discovery, category, comparison, recommendation and brand questions, written down and frozen so results are comparable month to month.

2. 2

   Establish a baseline across surfaces

   Run the set on the AI search surfaces available to us and record mention, recommendation, citation, source used, competitor mention and factual accuracy.

3. 3

   Identify which sources are being used instead of you

   This is often the most actionable finding: a directory, a review site or a competitor blog is answering questions about your business.

4. 4

   Audit entity clarity on your own site

   Can a machine extract who you are, what you do, who you serve, where, how you price and who works there — in plain sentences, without inference?

5. 5

   Check consistency off-site

   Website, LinkedIn, directories and profiles should state the same core facts. Contradictions across sources make every source less usable.

6. 6

   Check technical access

   Are the pages that explain your business crawlable and renderable without JavaScript? Are relevant crawlers blocked in robots.txt by accident?

7. 7

   Assess citable material

   Original data, research or worked examples give these systems something to reference. Marketing copy gives them nothing.

What we would investigate

1. 1

   Confirm the conversion data is real

   Submit the forms ourselves and trace each one through GTM, GA4 and the CRM. A meaningful share of low-conversion cases are measurement problems.

2. 2

   Measure conversion by template, not site-wide

   A site-wide rate hides both the pages that work and the pages that are broken. Almost every real conversion problem is concentrated.

3. 3

   Check whether the next step matches the buyer stage

   “Learn more” on a page reached by a high-intent commercial search is a genuine cause of lost conversion — the CTA is the wrong commitment level.

4. 4

   Test the form and checkout on mobile

   Required fields, validation messages, error handling, tap targets, and whether spam filtering is silently discarding real submissions.

5. 5

   Compare against paid on the same pages

   If paid converts and organic does not on identical pages, the difference is intent — a traffic problem wearing a conversion costume.

6. 6

   Look at what the page does not answer

   Scope, price guidance, process, what happens next, proof. Buyers who cannot answer these leave and shortlist someone who told them.

What we would investigate

1. 1

   Check whether new pages are even indexed

   Publishing into a site that is not indexing new URLs produces exactly this pattern, and it is invisible unless someone looks at coverage by page type.

2. 2

   Look for cannibalisation

   Several pages competing for the same query keeps all of them mid-page. At scale this is systematic and fixable with a cluster model.

3. 3

   Compare demand against what you publish

   Content briefed from volume rather than buyer stage produces informational traffic and no commercial movement. That is a structure problem, not an effort problem.

4. 4

   Audit internal linking

   If links are generated by the navigation template rather than by page importance, the pages you care about never accumulate anything.

5. 5

   Check the technical floor

   Rendering, canonicals, status codes and crawl distribution. Content work underperforms until the technical floor is solid, and it is cheaper to check than to assume.

6. 6

   Verify the measurement

   Some stalled programmes are actually working and being measured badly — brand and non-brand averaged together is the most common version.