Triage before treatment
Organic Traffic Dropped? File the Incident Before You Panic
Organic traffic dropped 40% overnight and the channel dashboard is red. Before anyone rewrites a title tag, you need two answers: is the loss real, and which clock are you on? This is the incident procedure we run for B2B and ecommerce sites, sorted into 48 hours, two weeks and one month.
- Real vs false drop
- 48h containment
- 2-week diagnosis
- 1-month recovery
No blame meetings. Log evidence first, hypotheses second.
§ 01 · Triage
First, decide whether the drop is real
Roughly a third of the “organic traffic dropped” alerts we are called about are measurement faults. Painting a real incident onto a false alarm wastes the most valuable days of the recovery window.
Probably a measurement fault
- Every channel falls at the same hour, not just organic
- Only one analytics property or one view is down
- Your tag manager changed, or consent mode went live
- Search Console shows flat impressions while analytics shows a cliff
- The drop lands exactly on a filter, a URL change or a migration date
Probably a real incident
- Clicks fall while impressions hold — you lost listings, not demand
- One template collapses while the rest of the site is flat
- Impressions fall with clicks, pointing to indexation or de-indexing
- Revenue falls in the same window through a separate pipeline
- The drop follows a core update or a link-related message in Search Console
The twenty-minute test: compare clicks and impressions separately, for one template and for the whole site, in both Search Console and analytics. If the two tools disagree on direction, fix measurement before you touch content.
§ 02 · Register
The incident register: what each signal usually means
Keep one register per incident. It stops the team from re-investigating the same hypothesis three times and gives you evidence if you ever need to explain the quarter.
| When | Signal | Probable cause | Action taken |
|---|---|---|---|
| Day 0 · 07:00 | Clicks −43%, impressions −6% | Listing loss on a template, not lost demand | Freeze releases, export 28 days of query-level data |
| Day 0 · 11:00 | Robots meta noindex on /products/* | Theme release shipped a staging flag | Revert header, re-crawl 50 URLs, confirm 200 + indexable |
| Day 1 · 09:00 | Coverage: 1,900 excluded, “blocked by noindex” | Global template inheritance | Request validation, monitor fetch rate daily |
| Day 3 · 09:00 | Clicks recovering 12% below baseline | Pages re-fetched, some still queued | Add internal links to the slowest template |
| Day 7 · 09:00 | Impressions above pre-incident level | Recovery, plus seasonal lift | Close tag, keep weekly coverage watch for 30 days |
| Day 14 · 09:00 | Inquiries back within 8% of plan | Demand intact, rankings restored | Post-mortem, add noindex check to release checklist |
§ 03 · Containment
The 48-hour clock: stop the bleeding
The first two days are not for strategy. They are for containment, evidence preservation and the three or four fixes that can be shipped today.
0–6 hours
Freeze and capture
Freeze. Halt content and template deploys. Ask engineering what shipped in the last 72 hours, including CDN and DNS changes, because the answer is frequently a header rule nobody classified as a release.
Capture. Export query, page and country data before anything changes. Screenshot the coverage report. Pull 14 days of server logs if you have them. Every later hypothesis gets tested against this snapshot, and without it you will argue about what “normal” looked like.
6–24 hours
Test the cheap hypotheses
- Indexability: fetch five revenue URLs as Googlebot, check status, robots meta, X-Robots-Tag and canonical.
- Manual actions: open the Search Console manual actions panel. A real message changes the entire response plan.
- Template scope: compare the affected pages against a template that still performs. Differences in header, canonical or rendering usually surface in minutes.
- Tracking: compare analytics against raw server requests or a log-based counter to rule out a tagging failure.
24–48 hours
Ship only reversible fixes
Revert the noindex, remove the blocking header, fix the redirect loop, restore the sitemap. Ship the reversible fix first and the optimisation later. Request validation where the console allows it, then re-crawl a sample and confirm the response actually changed — a deploy that did not clear the CDN cache has fooled more teams than any algorithm update.
Do not rewrite content in this window. You cannot know yet whether the problem is technical, and rewriting removes the evidence you will need.
§ 04 · Diagnosis
The two-week window: name the cause
If the loss survives containment, it is structural or algorithmic. Two weeks is enough to separate the five causes that matter, and too short to guess.
Sort by shape of loss. A cliff on one date with a broad recovery is usually technical. A staircase over several weeks is usually algorithmic. A slow slope across templates is usually competitive or a SERP layout change. A drop confined to one country or one device is usually infrastructure, hreflang or rendering.
Sort by template. Group URLs by template and compare click curves. If 80% of revenue lives on one template, protect it; if the loss is evenly spread across every template, look at site level: crawl budget, Core Web Vitals, a site-wide quality judgement, or a domain-level migration issue.
Sort by query intent. Lost branded clicks mean something different from lost non-branded clicks. Branded loss points to a technical or reputation problem; non-branded loss points to ranking, relevance or the SERP itself.
Then read the update calendar. Map the drop date against known core updates and spam updates. Correlation is not proof, but a drop that starts within 48 hours of a documented update, across unrelated templates, with no technical change, is unlikely to be a coincidence. The remedy for an algorithmic loss is quality and relevance work, not a resubmission request.
By the end of week two you should be able to write one sentence: “We lost X% of clicks on template Y, starting on date Z, because of cause C, evidenced by E.” If you cannot fill in C and E, you are not ready to spend a month fixing anything.
§ 05 · Root causes
Six causes behind most traffic drops
Each has a different recovery clock. Confusing them is why teams spend a month on the wrong remedy.
Core or spam update
Broad, template-agnostic loss within days of a documented update. Recovery comes from quality, originality and relevance — and it can take two to six months. No request form helps.
Manual action
A message in Search Console, plus a sharp site-wide or section-wide drop. Needs cleanup, documentation and a reconsideration request. Never a content refresh.
Technical accident
Noindex, robots block, redirect chain, expired certificate, 5xx storm, broken canonical. Detectable in hours with a fetch test. Fastest to fix and fastest to recover.
SERP layout change
AI answers, shopping units, local packs or video blocks push organic results down. Impressions hold, clicks fall, average position barely moves. Fix by earning the new surface, not by chasing rank.
Seasonality and holidays
Same weeks every year, visible in last year’s curve. Compare year over year, not week over week. The only real work is forecasting and protecting the peak.
Tracking failure
Tag manager edits, consent banners, cookie expiry, a new analytics property. Clicks are fine; the measurement is not. Verify with a raw log or a second source before escalating.
§ 06 · Recovery
The one-month plan: rebuild, do not just repair
Once the cause is named, the month is about restoring the pages that earn money and making the next incident cheaper to detect.
Re-index the priority set
Submit clean sitemaps, rebuild internal links to the affected template, and verify rendered HTML on the ten highest-revenue URLs. Track fetch rate daily.
Close the gap that caused it
If quality was the cause, consolidate thin pages, add first-hand detail and cite sources. If coverage was the cause, delete or redirect what should never have been indexed.
Instrument the next one
Alerts on coverage deltas, noindex counts, 5xx rate and click deviation by template. An incident that gets caught in six hours costs a fraction of one caught in six days.
A full SEO Audit Services review is usually how the month ends: a documented baseline so the next drop can be compared against a known state instead of a memory.
§ 07 · Evidence
What recovery looked like in three files
Sector descriptions, client-owned data, no brand names. Timelines are measured from the date of the drop, not the date we were hired.
Laboratory supplies distributor
Seven weeks after a core update, clicks returned to 94% of the previous level. Organic inquiries moved from 6 to 17 per month by month six.
B2B textile exporter
A theme release shipped a noindex across 1,900 product pages. Reverted in a day, re-indexed over nine days. Inquiries recovered from 4 to 12 per month.
Industrial pump manufacturer
Lost 38% of clicks to a SERP layout change with no rank loss. Comparison and specification content recovered the clicks in five months; inquiries 5 to 23 per month.
§ 08 · Scope
What an incident review includes
We start with a fixed-scope review: the register above, populated from your data, plus a written cause statement and a prioritised fix list. Pricing is quoted against scope after we see the coverage export and the drop window — a number before that would be fiction.
The review covers: drop classification, template-level click curves, index and render checks on the affected URLs, manual action assessment, update-calendar mapping, and a 30-day recovery plan with owners. If the cause turns out to be tracking, we say so and the engagement ends there.
You bring Search Console access, analytics access, 14 days of logs and one person who knows what shipped. We bring the register, the evidence and the plan. Use the SEO Audit Checklist as the shared vocabulary if your team wants to run the first pass internally.
§ 09 · Questions
FAQ for a bad Monday
How long does recovery take after a technical drop?
If the cause is a noindex, a robots rule or a redirect, the fetch usually resumes within days and clicks recover over one to three weeks as pages are re-crawled. If the site was fully de-indexed, expect four to eight weeks for the bulk of the recovery, with the long tail later.
Can you tell me on day one whether it was an algorithm update?
Not honestly. We can tell you whether the shape fits, whether the timing aligns with a documented update, and whether any technical change could explain it. That is a strong working hypothesis within 48 hours, not a verdict. The two-week diagnosis is what turns it into a cause statement.
Should I delete pages that lost traffic?
Rarely as a first move. A page that lost rankings may still convert, still hold links, and still serve branded demand. Consolidate only when the page duplicates another, has no unique value and no external links worth preserving. Deletion is a last step, not a reflex.
Does resubmitting the sitemap fix a drop?
Only if discovery was the missing piece. If pages are crawled and excluded, a sitemap resubmission changes nothing. It is one of the cheapest actions available, so we do it early, but we never present it as the fix.
What if it turns out to be seasonality?
Then the review closes with a forecast model instead of a fix list. Year-over-year curves, lead-time shifts and holiday calendars explain a large share of November and January panics. Knowing that is worth the fee on its own.
Will you guarantee the traffic comes back?
No. We can restore indexability, fix rendering, remove toxic patterns and improve relevance. We cannot control how a search engine reassesses a site, and anyone who guarantees a recovery date is selling certainty they do not have.
§ 10 · Open a file
Send the drop date and the coverage export. We will classify it today.
One file, one call, one written cause statement. If it is a false alarm, you will know within a day and save the month.