Your investigation confirmed there is a problem. It didn't tell you how to fix it.
A ranking drop report that proves traffic fell is not an investigation. It is a diagnosis. It tells you what Google did. It does not tell you what to fix, in what order, or what specifically to change.
Most reports stop at findings. Your team leaves the debrief meeting knowing the cause and no clear next step. Every week without an action plan is a week the gap between your position and your competitors’ widens.
For technical teams
A ranking drop investigation runs ten structured phases. Each must be consciously ruled in or out with recorded reasoning — not assumed unlikely and skipped.
Phase 01
Scope & data integrity
Confirm the drop is real in GSC actual clicks before investigating anything. Third-party estimated-traffic models have overstated a “drop” by more than 30 percentage points on real client accounts. Define a clean 14-day pre/post window, skipping the change week to reduce noise.
Phase 02
Timeline and trigger correlation:
Overlay the drop date against Google algorithm updates, on-site changes (H1/title/template/migration), and seasonality (check year-on-year). The drop date is the most powerful diagnostic clue available. Separate distinct events — an algorithm hit and an on-site change weeks apart are two causes, not one.
Phase 03
Segmentation:
Localise the loss by URL, subfolder, query, country, device, and crucially branded vs non-branded. “Traffic is down 30%” is not a finding. A branded-search spike can mask a complete topical collapse in the site-wide totals.
Phase 04
Technical & indexability:
Index status, canonicals, hreflang, robots/noindex, redirect chains, sitemap, schema. Pull per-URL detail from the audit API — dashboard CSV exports give summary counts only and will miss per-URL specifics.
Phase 05
Content & E-E-A-T:
Topical authority, semantic duplication, thin or AI-indistinguishable content, author and entity signals, freshness. Most core-update and Helpful Content losses sit here. Near-duplicate clusters and missing author credentials (YMYL) are the most common lever.
Phase 06
Rankings, SERP & intent shift
Position changes on target queries, new SERP features (AI Overviews, answer boxes, carousels) that eat clicks without changing rank, and intent shifts where Google now rewards a different page type for the same query.
Phase 07
Backlinks & link quality
Lost or gained referring domains, toxic-link spikes, anchor-text ratios. Assess quality, not count. If recommending a disavow, the toxic domains must be ranked by spam score — never hand over an unranked dump.
Phase 08
Core Web Vitals
Pull CrUX p75 LCP/INP/CLS as a confounder check, not a headline finding. CWV rarely causes a sudden ranking cliff. The key diagnostic question: did CWV actually move during the drop window?
Phase 09
GEO & AI Overviews
A modern loss vector. A page can hold its rank and still lose clicks because an AI Overview now answers the query above it. Check AI-search visibility for the affected queries and whether the brand is cited in AI answers.
Phase 10
Synthesis & root cause
Pull all evidence into a confidence-rated root cause. Classify: penalty (rare, confirm in GSC manual actions), algorithm, self-inflicted (on-site change), tool artifact, or competitive/SERP shift. State confidence and what additional evidence would change the conclusion.
WLDM delivers a prioritised task list. Not a report.
What you receive
Slide deck with findings
Prioritised task list ready to action
Recommendation format
Themes and observations
Exact URL, fix steps, owner, effort, and expected impact
Quality standard
Completeness of findings
Whether another team member can act on every task without asking a single question
What happens next
Debrief meeting
Your team starts on Monday
For technical teams
Before any WLDM investigation leaves the team, it is measured against one test: “If I handed this to another WLDM team member who has never seen this client, could they action every finding without asking me a single question?” If the answer is no, the investigation is not finished.
This test catches three failure modes that appear in most agency deliverables: findings described as themes rather than tasks, recommendations that name a problem without naming the fix, and action items with no URL, no owner, and no priority.
Every recommendation in a WLDM investigation is a row in a single prioritised table. No recommendation ships without all six fields:
Priority
Action (with exact URL)
How (steps / suggested copy)
Impact
Effort
Owner
P1
CRITICAL
Restore the topical H1 on /accounting-software/for-small-business
Set H1 to “Accounting Software for Small Businesses” — was changed to a generic H1. Single highest-leverage change.
Primary recovery lever; reconnects topical cascade.
S
DEV
P2 HIGH
Fix hreflang on the 579 pages pointing to non-canonical URLs
Point hreflang to canonical URLs; add missing self-reference and x-default. Exact URLs in the linked tab.
Stops wrong-region pages outranking the target.
M
DEV
P3 MEDIUM
Add keyword-rich internal links from /accounting hub to child page
Replace generic “learn more” anchors with “accounting software for small business”. 8 link sites in the linked tab.
Amplifies the H1 fix; passes hub authority down.
M
Content
P4 LOW
Disavow 12 toxic domains (ranked list linked)
Upload ranked disavow file to GSC. Low confidence this caused the drop — do last.
Hygiene; unlikely to move rankings alone.
M
SEO
Priority tier definitions: P1 CRITICAL — root-cause fix or highest-leverage quick win. P2 HIGH — strong lever, expected movement in weeks. P3 MEDIUM — supporting work that compounds the primary fix. P4 LOW — hygiene or low-confidence items. Rows ordered by impact ÷ effort; root causes always come before their downstream symptoms.
The most common failure in ranking investigations: a theme that looks like a recommendation.
“Consolidate semantically similar pages” is a theme. Turned into a task it becomes: “P1 — Merge the 3 apology-letter pages into one pillar at /blog/apology-letter-to-court, 301-redirect the other two, add ~300-word ‘Assault’ and ‘Drink-driving’ specifics sections. Owner: Content. Effort: M. Impact: recover ‘sample apology letter to court’ (fell position 1.5 → 20).”
The test: if a team member could not start work from this row tomorrow morning, it is not a task yet.
Five steps. Every investigation.
No cause ruled out without evidence.
Verify
We confirm the drop is real in actual performance data before investigating anything. Third-party traffic estimates have been wrong before — sometimes by more than 30 percentage points.
Locate
We find where exactly the loss is — by page, by query, by country, and by device. “Traffic is down 30%” is useless. We need to know which pages, which query groups, and whether the loss is concentrated or spread.
Timeline
We map when the drop happened against algorithm updates, on-site changes, and seasonal patterns. The date of the drop is the most important clue. We do not conflate separate events.
Diagnose
We work through every plausible cause — technical, content and authority, link profile, and SERP shifts — until we can name a root cause with a confidence rating.
Deliver
Every finding becomes a task. Named URL, exact fix, priority tier, effort, expected impact, and owner. Nothing in the deliverable is a theme or a suggestion.
For technical teams
Each phase of the diagnostic framework maps to specific data sources. Every tool call is documented in the final methodology section so every claim is auditable.
Phase
What is checked
Primary tools
1 — Data integrity
GSC actual clicks vs tool estimate; tracking gaps; analysis window
gscServer: compare_search_periods · get_search_analytics · ga4: list_property_annotations
2 — Timeline
Daily GSC trend; Ahrefs traffic history; Wayback page state
gscServer: get_search_analytics (daily) · ahrefs: site-explorer-metrics-history
3 — Segmentation
Clicks by page, query, country, device, branded vs non-branded
gscServer: get_search_by_page_query · get_advanced_search_analytics · ahrefs: site-explorer-pages-by-traffic
4 — Technical
Per-URL indexing, canonicals, hreflang, redirect chains, schema
ahrefs: site-audit-page-explorer · gscServer: inspect_url_enhanced · dataforseo: on_page_instant_pages
5 — Content / E-E-A-T
Duplication, thin content, author signals, freshness
seo-content · dataforseo: content_analysis_summary · ahrefs: site-explorer-top-pages
6 — Rankings / SERP
Position changes, new SERP features, intent shift
gscServer: get_advanced_search_analytics · ahrefs: serp-overview · dataforseo: serp_organic_live_advanced
7 — Backlinks
Referring domain changes, toxic links, anchor ratios
ahrefs: site-explorer-referring-domains · site-explorer-anchors · dataforseo: backlinks_bulk_spam_score
8 — Core Web Vitals
CrUX p75 LCP/INP/CLS; confounder check only
pagespeed: crux_query_history · psi_analyze_url
9 — GEO / AI Overviews
AI Overview exposure; brand citation in AI answers
seo-geo · dataforseo: ai_optimization_llm_response · ahrefs: brand-radar
10 — Synthesis
Root cause classification; confidence rating
Cross-phase evidence review
From 5.5 million visits to 12.8 million. After a core update cut traffic in half.
A large games portal was hit by a core algorithm update. Traffic fell roughly in half. WLDM diagnosed the root cause, built the recovery plan, and executed the authority engineering to back it.
Referring domains grew from month one. Traffic turned at month two. The recovery has held for over two years without continuous intervention.
+127%
Organic traffic from the post-update trough — 5.5M to 12.8M monthly visits.
Month 2
Referring domains moved first; traffic followed at month two. That lag is the mechanism proof.
27 months
The recovery has held as of June 2026.
For technical teams
The site (large puzzle and card games portal) was hit by a core algorithm update in mid-2023. Traffic declined from 7.2M in June 2023 to 4.4M by October 2023 — a 39% fall. By March 2024 it had stabilised at 5.5M with 3,148 referring domains. WLDM engaged March 2024.
Month
Organic visits
Referring domains
Note
2023-06
7.2M
2,562
Pre-drop peak
2023-07
5.6M
2,540
Core update impact
2023-09
4.9M
2,643
2023-10
4.4M
2,657
Trough
2023-12
5.1M
2,764
2024-03
5.5M
3,148
WLDM engaged
2024-04
5.6M
3,341
2024-05
8.3M
3,463
Traffic turns — +51% month-on-month
2024-06
7.8M
3,740
2024-08
8.8M
4,010
2024-09
10.1M
3,936
2024-12
11.3M
3,967
2025-01
12.8M
4,148
Peak
2025-02
12.8M
4,267
2025-06
12.5M
4,871
2026-06
9.9M
7,874
Recovery holding, 27 months on
The lag is the mechanism proof. Referring domains climb from month one (Mar→Apr, 3,148→3,341). Traffic turns at month two (Apr→May, +51%). That 4–8 week lag is the signature of real links being crawled, indexed, and processed through Google’s standard ranking pipeline: crawl latency (~4 weeks), ranking-cycle time (2–4 weeks), statistical signal accumulation. A link that moved traffic immediately would be a signal Google would discount. The lag is what makes the pattern trustworthy and repeatable.
The recovery has held. As of June 2026 — 27 months after engagement — the site holds 9.9M visits and 7,874 referring domains. Authority compounded throughout; the site did not require continuous intervention to maintain the recovery.
Source: case-studies-corpus.md §5 Case A; Ahrefs Site Explorer.
We will investigate the drop and deliver a task list your team can action this week.
Whether it happened last month or six months ago, the cause is findable. The action plan is what most investigations skip.
We run the full diagnostic — ten phases, cross-validated against real performance data — and deliver a prioritised task list with exact URLs, exact fixes, and a projected recovery timeline. We work directly with your team. We know the board is not waiting — and neither do we.
Free initial consultation. We will tell you what we can find before you commit to anything.
For technical teams
The investigation deliverable is a branded HTML report with nine required sections. Here is what each section contains.
01
Executive summary
2–3 findings in plain English a non-technical board member can understand. A metric grid of headline numbers (traffic, keyword, referring-domain deltas), a 0–100 weighted scorecard across seven categories, and Top-5 Critical Issues + Top-5 Quick Wins.
02
Action plan
The prioritised task table, placed directly after the executive summary. Every row: exact URL, fix steps, priority tier (P1–P4), expected impact, effort (S/M/L), and owner (Dev/Content/SEO).
03
Evidence sections
One section per finding. Each contains a chart and a per-URL or per-keyword table with pre/post and delta figures. Every data point linked to its source.
04
Competitive landscape
SERP tables for the target queries: which domains rank, their estimated traffic, domain authority, and whether the client appears in the top 10.
05
Confounders
What was ruled out and why: Core Web Vitals, seasonality, tracking gaps, tool artifacts — explicitly stated so they aren’t re-raised later as alternative explanations.
06
Site-audit scorecard
Health score and issue counts by category, each row linked to the per-URL Google Sheet tab with the full issue list.
07
Projected impact
Ranged recovery estimate tied to specific work — e.g. “60–80% recovery over 6–12 weeks if P1–P2 ship, assuming no further algorithm changes.” Ranges, not promises.
08
Methodology & data sources
Every tool, property, analysis window, and crawl date. Every claim is auditable.
09
Recovery projection arc
Expected score progression from current state → critical fixes → full plan, with the traffic and ranking movement expected at each stage.
WLDM closes every investigation with a recovery projection: an honest, ranged estimate that sets realistic expectations for the board without overpromising.
Current state — the weighted scorecard score as it stands today
After P1–P2 (critical and high fixes) — expected score improvement and traffic delta if only the highest-priority tasks ship
After full plan — expected score and traffic if the complete action plan is executed over 90 days
Every projection is ranged (not a single number), tied to which specific tasks ship, and qualified against algorithm stability. The same format is used for every investigation so progress can be benchmarked quarter over quarter.
