The short version
- A traffic drop has six possible causes. Diagnosis means eliminating five, not guessing the first.
- Always ask first: did only you fall, or did the whole market? It changes the answer entirely.
- A drop on one URL is a page problem. Across the whole domain it is technical or algorithmic.
- Stable positions with falling clicks means the SERP changed, not your site.
- Without a baseline saved beforehand, diagnosis degenerates into a debate of opinions.
The call always sounds the same: "we've lost 40% of our traffic, what happened?" And the honest answer, at the moment the question is asked, is that nobody knows yet — because an organic traffic drop is not a diagnosis, it is a symptom. The same numbers can describe an algorithm update, a deployment error, a competitor who published something better, or simply August.
This article is the procedure we run, in the exact order we run it. It is not a checklist: it is a sequence designed to eliminate wrong hypotheses quickly, because the cost of fixing the wrong problem is higher than the cost of waiting two days to understand the right one.
Step 0: touch nothing
The instinct when traffic collapses is to do something. It is almost always the most expensive mistake of the week.
If you change content, canonicals, redirects and internal linking in the first 48 hours, you have introduced four new variables into a system you have not yet understood. When traffic moves the following week, you will not know whether it was one of your changes, natural recovery, or both pulling in opposite directions.
The rule is simple: measure first, then touch. The only exception is a confirmed, obvious technical fault — a noindex in production, a server returning 500s, a robots.txt block. Those get fixed immediately, because they are not hypotheses.
Asking three different consultants what to do and receiving three incompatible plans. Each will see the problem through their own specialism: the technical person finds a technical problem, the copywriter finds a content problem. The diagnosis has to come from the data, not from the profile of whoever is looking.
The six possible causes
Before looking at any graph, it helps to hold the full list in mind. There are six, and each leaves a different signature in the data.
| Cause | Signature in the data | Speed of the drop |
|---|---|---|
| Technical fault | Sharp fall, often site-wide, with a precise date | Sudden, within 24-72 hours |
| Algorithm update | Broad decline; competitors move too | Progressive over 1-2 weeks |
| Competition | Losses on specific queries, one competitor rising | Slow, over months |
| SERP layout change | Positions unchanged, clicks down | Sudden, with no ranking loss |
| Seasonality | Repeats in the same period last year | Predictable |
| Content decay | Slow decline on older pages, concentrated | Very slow, over quarters |
Note something: only two of the six require urgent intervention. The other four require understanding first, and in three cases out of four the correct response is to change the editorial plan rather than to rush at the website.
The diagnostic sequence
- Is it actually a drop?Compare against the same period last year, not against last month. In a city like Naples seasonality is violent in whole sectors: hospitality, restaurants, anything tied to tourism. A 35% fall in August against June can be a 10% rise against last August - in which case there is no problem to solve.
- Did everything fall, or part of it?Filter by URL and by query group. A drop concentrated on a handful of pages is a problem with those pages. A uniform site-wide drop is technical or algorithmic. This single distinction eliminates half the hypotheses in ten minutes.
- Did positions actually change?Here rank tracking answers a question analytics alone cannot: if you lost clicks but positions are steady, your site did not get worse - the available space on the results page did.
- Did the market move too?Compare against three direct competitors. If everyone fell, it is an update and the correct reaction is to wait two weeks. If only you fell while one rose, you have your own problem and it is worth seeing what they published.
- Is the site still reachable and indexed?robots.txt read in full, status codes on the main URLs, and above all the bot visit log: if crawlers stopped visiting a section, you have found the date and the exact location of the problem without having to infer it.
- What changed on your side?Deployments, plugin updates, migrations, redesigns, hosting changes, menu edits. Nine times out of ten the answer is here, and nine times out of ten nobody connected it to the drop because "it was only a design change".
"What went live in the fifteen days before the drop?" Ask whoever manages the website, not whoever manages marketing. The useful answers nearly always come from there.
When it is an algorithm update
Updates generate more panic than they deserve, because the vocabulary around them is dramatic and the practical guidance is thin.
How to recognise one: the decline is progressive over one or two weeks rather than sudden; it touches many queries rather than a few; and above all the whole SERP moved, not just you. If you check your three competitors and see that two fell and one rose, you are not looking at a technical fault of your own.
What to do, in order: wait two weeks, because updates have settling periods and intervening during a rollout means correcting something that would have corrected itself. Then analyse which pages lost and what they have in common — a pattern almost always emerges: thin content, pages with no differentiation, topics where you have no direct experience to show. Then work on that pattern, which is months of editorial work rather than a week of fixes.
What not to do: change everything at once. If you alter content, structure and internal links simultaneously and recover three months later, you have learned nothing reusable.
When the SERP changed (the case nobody diagnoses)
This scenario deserves its own section because it is systematically mistaken for a ranking loss, and the two require opposite actions.
The picture: your positions are identical to two months ago, impressions are flat or even up, but clicks are down 30%. You have not lost ranking. You have lost space. Something appeared above your result that was not there before: a local pack, a generated answer block, more ads, a carousel.
For a local business this happens most often on commercial local queries, where the three-listing pack occupies the dominant visual position and organic results slide below the fold.
The correct response is not to rewrite the page. It is a choice between three routes: compete inside the block that took your space, if it is a local pack the listing is workable; take the block itself, if it is a featured snippet; or accept that the query has become less profitable and move the investment to queries where organic results are still visible.
The most common obstacle: nobody saved the "before"
The most frequent practical difficulty is not identifying the cause. It is that the data needed no longer exists.
Search Console retains sixteen months of history. If the decline started fourteen months ago and you notice now, the year-on-year comparison is already partly gone. If the site changed property, domain or monitoring tool, the "before" may have disappeared entirely. And positions are stored nowhere at all unless somebody was tracking them.
Which is why the most useful part of this article is not about drops but about prevention: keep your own baseline, current, in an account you control. Three things are enough, all available in the free layer of Semalt — clicks and impressions per URL, positions by query group, and an indexing log — and their value is not daily. It is the day something happens.
If the numbers come from the supplier being evaluated, every discussion turns political. With an independent source, "what happened?" goes back to being a technical question with a verifiable answer.
What to communicate while investigating
An overlooked aspect: during the diagnosis somebody inside the business is anxious, and an information vacuum gets filled with hypotheses worse than reality.
Three sentences resolve it almost every time. What we know: the drop started on this date and affects these pages. What we are ruling out: we are checking whether it is technical, algorithmic or market-driven, in that order. When we will have an answer: a specific date, typically within seventy-two hours.
What to avoid is generic reassurance — "it's normal, it happens" — which is neither informative nor, often, true.
A protocol worth having ready
The difference between businesses that handle a drop well and those that go into crisis is not technical skill: it is having decided in advance what to do. Half a page, written while everything is fine:
Who looks first and within how many hours. Which five checks they run, in what order. Who decides whether to act or wait. What must not be touched during the investigation. Who informs the rest of the business, and how often.
It sounds bureaucratic and it saves weeks, because it eliminates the most expensive phase of any incident: the first three days when everyone is doing something and nobody knows what.
Save your baseline today
Clicks per URL, positions by query group and a bot log in an account you own. Free, ten minutes, and it earns its keep on exactly the day something goes wrong.
Sign in to Semalt See the analyticsFrequently asked questions
How long should I wait before acting?
Two to three days for diagnosis, and two weeks before reacting to a suspected algorithm update. The exception is a confirmed technical fault, which gets fixed immediately.
How do I know it is an update?
The decline is progressive, touches many queries, and competitors move too. If only you fell, it is not an update.
I lost clicks but not positions. Why?
The SERP composition changed: new blocks above you reduced visible space. Rewriting the page will not help; deciding whether to compete for that block or reallocate the investment will.
The drop looks seasonal - how do I prove it?
Compare with the same period last year. If the profile repeats, it is seasonality and should be planned for rather than corrected.
I have no historical data. What now?
Reconstruct what you can from Search Console's sixteen months and start keeping a baseline today. The next drop will be diagnosable in hours rather than weeks.
Conclusion
A traffic drop should be handled like a fault, not like an emergency: isolate the variables, rule out causes one at a time, and act only once you know what you are acting on. The temptation to move immediately is strong and is almost always the reason why, three months later, nobody can explain what happened.
The most useful thing you can do today, while nothing is wrong, is prepare for the day something is: connect the domain and store clicks per URL, positions and indexing status. Ten minutes now is worth weeks of avoided argument later.
And if the drop is already happening and you want a methodical diagnosis, get in touch: the initial audit is free, delivered within 24 hours, and in most cases the cause emerges in the first two checks.
