By situation
Something changed. Establish what before deciding what to do about it.
The instinct after a decline is to act immediately, and acting on the wrong theory costs more than the week spent being sure. Most failed recoveries begin with a confident diagnosis made in the first forty-eight hours.
If a migration launched recently, treat this as urgent — the recoverable window narrows week by week.
A decline in search traffic produces a predictable sequence. Somebody notices. A theory is proposed, usually an algorithm update, because it is the explanation that requires no one to have made a mistake. Work begins against that theory. Traffic does not recover, because the theory was wrong.
The alternative is unglamorous and considerably more effective: establish precisely what fell, precisely when, and what changed in the same window. That evidence usually narrows the cause to one or two possibilities within a few days.
Diagnosis is urgent because the evidence degrades. Response is not urgent, and should wait until the diagnosis is sound.
Diagnosis
Establishing what actually happened
Six steps, in order. Most declines are attributable by the fourth.
Confirm the decline is real
Rule out measurement change first. A consent banner, a tag deployed incorrectly, a filter added to a view or a switch between analytics versions all produce a step change indistinguishable from lost traffic. Cross-check against Search Console, which measures differently.
Date it precisely
A gradual erosion over three months and an overnight fall on a specific date have entirely different causes. The shape of the decline is diagnostic in itself, and imprecise dating is the single most common reason a diagnosis goes wrong.
Establish what fell
Which pages, which searches, which devices, which regions. A uniform decline across everything suggests something site-wide. A decline confined to one section suggests something specific to it. This distinction eliminates most candidate causes immediately.
List every change in the window
Deployments, content changes, platform updates, plugin changes, server migrations, DNS edits, redirect rules. Include changes nobody thought were relevant, which is where the cause is most often found.
Check the external context
Confirmed algorithm updates and their dates. Competitor movement. Seasonal patterns in your market. Changes in how results are presented for your searches, which can reduce clicks without any change in position.
Form and test a hypothesis
Reach a specific account of the cause that is consistent with all the evidence, and identify what would confirm or refute it. Only then commit to a response.
The first explanation offered is almost always an algorithm update, because it is the only explanation in which nobody made a mistake.
Causes
What declines usually turn out to be
In rough order of frequency across the diagnoses we conduct.
A migration or redesign
By some distance the most common. Frequently launched weeks earlier, with the decline attributed to something else because the site 'went live fine'. Incomplete redirects and lost content are the usual mechanisms.
A technical change nobody flagged
A directive added during maintenance, a plugin update altering how pages render, a caching change, a certificate problem affecting a subdomain. Small, invisible, and entirely sufficient to cause a substantial loss.
A measurement artefact
Not a decline at all. Consent changes, tagging errors and analytics migrations produce a step change that looks identical to lost traffic and requires no remedy beyond correcting the measurement.
Competitive displacement
A competitor published something genuinely better, or a new entrant took positions you held. Gradual rather than sudden, and visible in the pattern: specific searches rather than everything.
An algorithmic reassessment
A real cause, and less common than assumed. Confirmed by timing that aligns precisely with a known update and by a pattern consistent with what that update addressed.
A change in result presentation
Position unchanged, clicks reduced, because the results page now answers the question directly or gives more space to something above you. Frequently not recoverable, and worth knowing rather than working against.
Shape
What the shape of a decline tells you
Before any investigation of causes, the pattern of the loss eliminates most of them. This is the cheapest diagnostic work available and it is routinely skipped in favour of proposing a cause.
A sudden fall on a single date is something that happened on that date — a deployment, a configuration change, a launch. A gradual erosion over months is competitive or reflects slow reassessment. A decline confined to one section is specific to that section. A decline across everything is site-wide.
Combining the shape with the date narrows most declines to one or two candidates before anybody has examined a single page.
Overnight, everything
A technical or configuration change. Look at what was deployed that day.
Overnight, one section
Something specific to that section — a template change, a directive, a restructure.
Gradual, specific searches
Competitive displacement. Somebody is doing better on those particular subjects.
Gradual, everything
Broad reassessment, or an accumulation of small changes over a long period.
Clicks down, positions steady
How results are presented has changed, not how you rank.
Frequently asked questions
How quickly should we act?
Everyone says it was an algorithm update.
Could this be a tracking problem rather than a real decline?
We changed several things at once.
Can traffic be recovered?
Establish the current state
The snapshot gives you an evidenced picture of the site as it stands now, which is the baseline any diagnosis works from.