Analysing a traffic drop
A traffic drop has several documented causes: an algorithmic update, a technical issue, a security alert, a manual action, seasonality, a site move, or a reporting glitch. The documentation recommends comparing over sixteen months before concluding, and advises against radical changes to a page that already performs.
A falling curve does not say why
This is the situation that produces the most hurried decisions: traffic falls, somebody asks what we are going to do, and an answer arrives before the diagnosis. The published documentation devotes a whole page to the question, and its first contribution is to recall how many different causes produce the same curve.
The documentation lists the causes to examine when traffic drops: an algorithmic update, technical issues, security issues, spam issues that can lead to a manual action, seasonality and changing interests, site moves, and a data processing anomaly.
officialSource : Google Search Central, “Debugging drops in Google Search traffic”
Seven families, two of which have nothing to do with the site — seasonality, which belongs to the audience, and the processing anomaly, which belongs to the measuring tool. Put another way, two times out of seven the right action is to do nothing, and no amount of content review would have revealed it.
The two checks that cost two minutes
The documentation recommends extending the displayed period to sixteen months before drawing any conclusion, to check that the drop is not the one that returns every year on the same dates.
officialSource : Google Search Central, “Debugging drops in Google Search traffic”
The documentation shows the textbook case: a curve diving exactly as it did the previous year, in the same week. An agency called in urgently over that curve will bill a rebuild for a phenomenon that will repeat next year, identically, whatever it does.
The second check is of the same kind: the data anomalies page, which lists measurement and logging incidents. A drop that exists only in the report has no remedy other than patience.
The documented causes are best checked in order of how much the check costs: reporting anomaly and seasonality first, security and manual actions next, rewriting content last.
hypothesis
That order is ours: the documentation lists the causes without ranking them. The order we propose does not follow likelihood, which no source publishes, but the cost of checking. Ruling out a cause in two minutes before exploring one that will take two weeks commits you to nothing.
What the shape of the drop indicates
If impressions stay flat while clicks fall, the documentation points to the title and snippet shown, or to a more appealing competing result, rather than to a ranking problem.
officialSource : Google Search Central, “Debugging drops in Google Search traffic”
Here is the most useful sorting rule on the whole page, and it fits into one comparison. Flat impressions mean the pages still appear as often: the problem is downstream, in what the reader reads before clicking. Impressions falling together with clicks send you back to the full list of causes instead.
The documentation then suggests isolating the scope: compare two periods, sort the pages by click difference, and see whether the loss is spread across the whole site or concentrated in a group. A general fault points to indexing, a local one to inspecting addresses.
Your traffic dropped 40%: you were hit by the latest update. Our recovery plan rewrites your top thirty pages and restores your positions within sixty days.
What the documentation refuses to let you believe
A drop following an algorithmic update does not imply a defect in the content: the documentation writes that there might not be anything fundamentally wrong with it.
officialSource : Google Search Central, “Debugging drops in Google Search traffic”
That sentence is worth knowing for any publisher who has lived through a major update. It does not say the content is good, it says the drop does not prove it is bad: the web changes, other pages arrive, and a position moves without any fault having been committed.
The documentation advises against radical changes to a page that is already performing well, and notes that small shifts in position happen at any time, including upwards and without any action.
officialSource : Google Search Central, “Debugging drops in Google Search traffic”
The text separates two situations explicitly: going from second to fourth place, which is ordinary variation, and falling out of the top ten across a wide range of terms, which calls for a review of the whole site. Confusing the two leads to rewriting a page that had nothing wrong with it.
After a change, some effects show within days and others take several months; the documentation suggests waiting a few weeks before analysing again, and notes that no change guarantees a noticeable impact.
officialSource : Google Search Central, “Debugging drops in Google Search traffic”
Establishing the cause of a drop before committing to a fix
hypothesis
- Effort
- low
- Time to effect
- immediate
- Risk
- none
- Durability
- structural
The effort is the lowest on this site: it means reading reports that already exist. The delay is immediate, since the result of the analysis is available the same day. The risk is nil, and this is the only card on this site where doing nothing is sometimes the correct conclusion. Durability is excellent: the procedure is reused at every incident.
What does not work, or no longer works
Urgently rewriting the pages that lost the most, first. It is the most common and most expensive reaction, and it assumes an answer to the question nobody asked.
Then comparing your curve to a competitor’s as estimated by a third-party tool. Those estimates share neither the source nor the method of the console’s measurements, and a drop in an estimate is not a drop in traffic.
Finally, hunting for the update responsible before ruling out seasonality and the reporting anomaly. An update date can always be found somewhere, and it will explain any curve to whoever goes looking.
Sources
- Debugging drops in Google Search traffic — Google Search Centralaccessed on
- Google Search spam policies — Google Search Centralaccessed on
Block revalidated on
Frequently asked questions
- Where do you start when the curve falls?
- With the period displayed. The documentation recommends switching to sixteen months, because a drop that returns every year on the same dates has no other explanation. Next comes the data anomalies page, which lists reporting incidents. Those two checks cost two minutes and rule out two whole causes.
- Clicks are down but impressions are flat, what does that mean?
- The documentation points to presentation rather than ranking: the title and snippet, or a competing result that became more appealing. The pages still appear as often, they are simply chosen less. The fix then lies in what the reader reads before clicking.
- Does a drop after an update mean my content is bad?
- Not necessarily, and the documentation says so explicitly: there might be nothing fundamentally wrong to fix. The web changes constantly, and a position is not an asset. The text separates the small shift, which it advises leaving alone, from a clear fall out of the top results, which calls for a site-wide review.
- How long should you wait after a fix?
- A few weeks before analysing again, according to the documentation. Some effects appear within days, others take months. The same text adds the sentence no supplier quotes: nothing guarantees that a change produces a noticeable impact.
- How do you tell whether the drop hits the whole site or a few pages?
- By comparing two periods in the pages table and sorting by click difference. A fault spread across the whole site points to indexing; a drop concentrated on a group of pages calls for inspecting addresses one by one. The shape of the problem says where to look.