Traffic drop diagnosis
Investigation to find the real cause of a traffic loss, ruling out hypotheses one by one instead of blaming the latest update.
Request an initial diagnosis- 15
- years diagnosing traffic drops
- 6
- senior specialisms inside the team
- 48 h
- to reply, at most
Brands we've worked with
Do any of these situations sound familiar?
A drop isn't explained with a hunch. It's explained with a date, a data point and the hypotheses you've been able to rule out.
The site is falling and you don't know why
The obvious suspicions have been checked, a couple of things tried, and the curve hasn't moved. Nobody can point to the cause with data.
The site is falling and the team can't agree on why
There's a technical explanation, a content one and a market one on the table, and none holds up with enough evidence to decide.
The site keeps losing month after month and won't stop
It isn't a one-off step down: it's a slope running for months and none of the actions tried has changed it.
The site migrated at the same time as an algorithm change
Template changes, releases or a migration and an algorithm update within days of each other. Nobody has separated what did what.
The site suffered a sudden, sharp drop
Overnight a chunk of traffic is gone and nothing seems broken: no visible error, just the chart going down.
The site holds traffic but loses sales
Sessions hold but revenue drops, and the discussion has been going in circles between marketing, product and data for weeks.
Four fronts always reviewed
They're in every diagnosis, though the weight changes: a one-day drop after a release isn't the same as a slow six-month loss.
Timeline and scope of the drop
Know when it started, what it took down and which parts of the site are still intact.
See the detail
- Series rebuilt by page type, country and device
- Separating drops in impressions, CTR and conversion
- Comparison against seasonality and previous years
- Identifying the URLs that concentrate the loss
Technical and indexing causes
Rule out the reversible before looking for big explanations: there's usually a date and a release behind it.
See the detail
- Cross-referencing the drop with releases, template changes and migrations
- Review of indexing, canonicals, redirects and status codes
- Rendering and served content versus what the crawler sees
- Log and crawl review when access is available
Content, intent and competition
Understand whether demand changed, the SERP changed, or the content stopped answering the query.
See the detail
- Real category demand over time, not just your brand
- SERP changes: new formats, AI Overviews and who has moved in
- Cannibalisation and pages competing with each other
- Content quality and freshness versus whoever overtook you
Business and measurement
Check the drop is real and what it's worth, before building a plan around a data error.
See the detail
- Analytics validation: tagging, consent and configuration changes
- Impact on revenue and leads, not just sessions
- Business decisions that affected catalogue or offering
- Defining the starting point recovery gets measured against
A drop isn't guessed: it's ruled out hypothesis by hypothesis
First look at the curve's shape; then eliminate candidates in order of cost, starting with what's reversible. Whatever is ruled out is written down with the data that rules it out.
Measurement change
Tagging, consent or filters breaking the series.
Release or template change
A release date that matches the step down.
Indexing and crawling
Canonicals, noindex, redirects and status errors.
Algorithm update
Only confirmed if date and pattern match.
SERP change
New blocks, AI Overviews and new competitors.
Demand and seasonality
The whole category falls and your share hasn't changed.
Business decision
Withdrawn catalogue, pricing, offer or country changes.
Who does it?
Senior profiles who have seen many drops and know what to check first. Whoever investigates is who explains it and who reviews it again.
Why Laika?
Because a badly diagnosed drop gets paid for twice.
First in lost traffic and then in months of work aimed at the wrong place. Investigating properly costs weeks; getting it wrong costs quarters.
The problem first, solutions after
Defining actions without understanding what's happening means working at random. Here the problem gets identified first, with a date and data behind it, and only then do we decide what to do.
We document what we rule out
Knowing what it wasn't is worth as much as knowing what it was: it saves months of work aimed at the wrong place.
We separate the reversible from the structural
Some drops are fixed in one release; others require rethinking the architecture. We don't treat them the same.
We tell you when it won't recover
When demand is gone or the SERP has changed for good, we say it and propose where to put the budget instead.
Whoever investigates is who explains it
No middle layers: the senior profile reviewing the data is the one sitting with leadership and engineering.
We look at AI and search together
Today a drop can come from AI Overviews or generative answers. It's measured like any other hypothesis.
Frequently asked questions
What if the cause can't be fixed?
We'll tell you that too. Knowing it won't recover saves months of misplaced investment.
How long until I see results?
The first findings and decisions land within weeks. Impact on visibility and revenue depends on your platform and your team's implementation speed: on mid-sized projects we usually talk about three to six months for solid movement.
I worked with another agency and it didn't work. Why now?
What usually fails is judgement, not effort: tasks get executed without deciding what matters. Before proposing anything we review your data, and if the problem isn't visibility we'll tell you, even if it means not selling the project.
Tell us what happened
With the approximate date of the drop and what you've already checked, we can tell you where we'd start. We look at your project before proposing anything.
Lost traffic and nobody knows why?
We rebuild the timeline, rule out hypotheses with data and tell you what's recoverable and what isn't.
We look at your project before proposing anything. Reply within two working days at most.
Talk to us about your drop