Auditing after a traffic drop: first find out what exactly went down
A site audit after a traffic drop is an investigation with dates and numbers, not a list of hypotheses. First, we break down the fall: which section, which type of pages, which language version, which device and country, from which day. Then we check the date with two calendars - algorithm updates and your own history of changes on the site. At the exit, the reason, the evidence for it and the division into "we will refund", "we will partially refund" and "we will not refund". The first diagnosis is free.
from the drop layout according to the webmaster, not from crawling the site
after a free audit
Price
30
Guarantee
days after project sign-off
from the drop layout according to the webmaster, not from crawling the site
Where do we start?
5-25
Term of paid audit
working days, depends on the number of reasons
free fall diagnosis, 2-4 working days
In front of him
repeated measurements 4 and 8 weeks after corrections
Control points
the second, outdated site map parallel to the canonical one
A frequent find
returning traffic to previous figures
What we do not promise
Who it's for
Situations where this service delivers results
Scenario 1 of 5
The traffic stopped one day, and no one knows why
A sharp break almost always has a contour: one section, one language branch, one pattern. You can find this contour in an hour of working with the webmaster's office, and it is he who throws out of consideration most of the hypotheses that managed to be named during the week of panic.
We'll review your situation in a free audit
Scenario 2 of 5
The fall has been going on for months, and no one has recorded the beginning
Here it is more difficult: there is no cut-off date, there is a slow slide. We break it down by month and by type of pages separately - it often turns out that one group of pages has been falling for a long time, and the rest are holding, and the total number simply hid it from the eyes.
Scenario 3 of 5
One language version fell, the second is in place
Most often, this is not an algorithm, but a technical hole: the version fell off the site map, the hreflang was declared asymmetrically, the redirect to the language branch began to lead to the wrong place. Such things are checked in a day and fixed quickly - for a long time they are usually simply not noticed.
Scenario 4 of 5
There was a move, redesign, or template change before the fall
Then we compare not hypotheses, but addresses: which pages existed before, which exist after, which give 404 or 301 and where exactly. The most expensive mistake here is believing that you have "moved everything" without a list of old addresses at hand.
Scenario 5 of 5
The contractor said "it's the updated algorithms" and that was it
The wording is convenient because it is not checked and does not oblige anyone. We put your drop date next to your update dates and next to your releases - and it happens that the latter coincides with the former, not the former.
Cases
Tasks and results in numbers — all metrics measured by us
a wholesale store of cosmetics for the beauty industry, a catalog of about 4,000 items — our technical measurement of an operating store
Task
Find a technical reason why a smaller part of the directory is indexed than it should be.
Solution
We bypassed all the addresses of the site maps that the server provides, counted the entries and unique addresses separately for each, compared them with the live directory and language branches.
Result
The canonical map gives 4,167 addresses: 3,947 products, 166 brands, 47 categories, 6 information pages and the main one. At the same time, the standard address is out of date: 7,782 records, of which only 2,755 are unique, and 2,454 addresses are repeated three times, and about 2,706 against 3,947 commercial addresses. The crawler, which follows the standard path, receives both a duplicated and an incomplete directory. Separately: there is no address of the second language version in the maps, although the hreflang and language switcher on the site work. First server response 576.8 ms at 192.4 KB of HTML. Measured on 07/31/2026.
a brand store of a natural cosmetics brand for the EU market — our technical measurement of an operating store
Task
Find out what generally confirms the dynamics of store attendance.
Solution
We counted the catalog with a regular engine search, checked the service files and conducted a full inventory of the counters on the live pages.
Result
Catalog of 23 products — the entire assortment of the brand, no pagination. Three language versions bg/en/ru with hreflang, six types of structured data on master, first server response 512.6ms at 220.4KB HTML. At the same time, /sitemap.xml, /robots.txt and the engine feed return an empty body or markup instead of XML, and there is no counter on the main one - no tag manager, no gtag, no pixel. There are no external analytics here at all, so any talk about crashes is based only on the webmaster's data. Measured on 07/31/2026.
What's included
Complete list of work and what you get as a result
Breakdown by sections, page types, language versions, devices and countries - so that you can see the contour, and not the general figure "minus 40%"
Exact start date checked against two calendars: algorithm updates and your own release history
Separating drop in impressions and drop in conversions are different diagnoses that are treated by different jobs
Technical validation: indexing, sitemaps, redirects, canonical URLs, robots.txt - counting unique addresses, not trusting the number of entries in the file
Content check: merged pages, changed templates, removed blocks, cannibalized two of your pages in one request
Parsing the server logs when access is available is the only way to see a change in bot behavior rather than assume it
Reconciliation of dynamics with demand in the industry — so as not to fix what wasn't broken
The conclusion is divided into "we will return", "we will partially refund" and "we will not return, we continue to work", with an explanation of why each point is there
Fixes plan in order of impact, with 4 and 8 week checkpoints
Raw data with a date and described method - you can work with it, even if another contractor corrects it
When this service isn't right
What's not included — so there are no surprises at delivery
Promises to return traffic to the previous level - no one works like that
Work with manual measures without access to the webmaster's office
Compensation of seasonal decline in demand by means of optimization
Correction of what was found: these are separate works with a separate assessment after diagnosis
Process steps
Transparent stages with approval at every step
Total duration:5–25 days
1
Fall layout
1-2 working days
Comparison of periods in the webmaster's office: pages, requests, countries, devices, language branches. Here we separate the drop in impressions from the drop in conversions.
2
Chronology and two reconciliations
1-2 working days
We put the start date next to the algorithm update calendar and next to your history of changes: releases, template, engine, hosting, text edits.
3
Technical cut
2-5 working days
Indexing, all site maps with counting of unique addresses, redirects, canonical addresses, robots.txt, response codes for each language branch separately.
4
Content slice
2-5 working days
What has changed on the pages: consolidated sections, revised templates, removed blocks, two separate pages for one request. We check with archival pictures.
5
Logs and demand
1-4 working days
The actual behavior of the bot according to the logs, if the hosting stores them, and comparing the dynamics with the demand in the industry: some of the crashes have no reason inside the site.
6
Conclusion and plan
1-3 working days
Reason with evidence, division into reversible and irreversible, works in the order of influence, control points. Next, a conversation about the document together with your developer.
7
Hypothesis testing
in a separate step, after corrections
Repeat measurements after 4 and 8 weeks using the same method. If there is no movement, the hypothesis was false, and this is also recorded as a result.
Free fall diagnosis
The first thing we do is break down the fall into parts. Very often "traffic down" means "one section down" or "one language version down" and these are different reasons with different fixes.
What we measure
What exactly fellSection, page type, language version, device, country. A drop evenly across the entire site is rare — and it means the most serious reasons.
When it fellExact start date. We compare it with two calendars: algorithm updates and your history of changes on the site.
Impressions or clicksA drop in impressions and a drop in conversions for the same impressions are two different stories with different reasons.
Indexing statusHow many pages were removed from the index and in what status. Here we often find a technical reason that no one knew about.
Site map and language branchesClassic: only one language version got into the site map or two maps with different content live in parallel.
Is there anything to measure?The first diagnostic question is whether your analytics work at all. We met sites where there is no counter, and then the drop itself is not confirmed by anything.
What you get
Fall schedule: what, when and how much, in numbers from your cabinets.
The main hypothesis of the cause and what it is supported by.
A list of what can be returned and what cannot be returned.
Conversation for 30 minutes on the document.
Timeline: 2-4 working days
Why is it free
Because in half of the cases, the cause is found two days in advance and is treated with one correction. We will not sell a monthly investigation where the Ukrainian version simply stopped getting into the site map.
What's next
If there is only one reason, we evaluate the correction. If there are several reasons or the drop drags on for months, we agree on a paid audit with a breakdown by types of pages and logs.
Short form: your contact and site URL
Contract, act and 30-day warranty
Every project gets a written contract: scope, deadlines, amount, acceptance procedure. After delivery — act and invoice, then 30 calendar days of warranty.
Sole proprietor & bank transfer
The contractor is a registered sole proprietor. Payment by invoice with closing documents.
Rights & access — yours
Code, design and materials transfer to you after full payment. Domain, hosting, repository and analytics are registered to you.
Client portal instead of email chains
During the project you get access to a portal: contracts, invoices, acts and project status in one place.
European clients
Among our work — projects for Norway, Bulgaria, Moldova and Spain.
Verifiable numbers
Every case in the portfolio comes with a link to a live site and a technical measurement.
Audit first, then pricing
There is no price list on the site intentionally: the scope of the same work differs multiples between clients.
We say "no" when unsure
If the task isn't ours or the deadline is unrealistic — we tell you upfront.
Did not find your case?
Describe how it works on your side — we will tell you whether “Audit after a traffic drop” fits and what it means in your situation. No brief and no call: one question, one answer.
What affects the price
Why two seemingly identical tasks are priced differently
How many reasons reallyOne technical reason is two days away and is covered by free diagnostics. Two or three superimposed on each other are separate lines of evidence, which also have to be untangled by dates.
The depth of the fall storyLast week's precipice is analyzed based on fresh data. The slide, which stretches for a year, rests on the limit of Search Console: there lie 16 months, then the picture has to be restored with archival images and your releases.
Size and heterogeneity of the siteA hundred pages are laid out in an hour. A catalog of tens of thousands of addresses has to be divided into types of pages and counted separately, because in the sum they hide each other.
Are the server logs available?The logs show how the bot's behavior changed, not how it could have changed. Without them, part of the conclusions remain assumptions, and we sign them just like that.
Language versions and subdomainsEach branch is decomposed separately: in a bilingual store, one version may fall, but in the overall figure it looks like a moderate fall of the entire site.
Are control measurements required?Hypothesis testing after 4 and 8 weeks is a second cycle of work using the same method and on the same pages. The audit itself does not cover it, and it is better to agree on this before the beginning.
Technologies & integrations
What we build on and what it connects to
Stack
Search Console — comparison of periods by pages, queries and countries; limit: only 16 months of history
GA4 — behavior on the site next to impressions; limit: blockers and consent mode make it a trend
bypassing all site maps with a count of entries and unique addresses; limit: addresses in the map are not equal to pages in the index
Screaming Frog — structure, response codes, canonical; limit: shows the site today, and the drop started earlier
server logs — the actual behavior of the bot; limit: hostings often hold them for 7-14 days
archival snapshots of pages — what has changed in the template; limit: the pictures are irregular, small corrections are not visible in them
checking response codes and redirect chains for each language branch separately
Integrations
Google Search Console
GA4
Bing Webmaster Tools
Cloudflare
Investigation by data against checklist-audit
How this option differs from the alternative
General SEO audit or auto-report of the serviceOur approach
Where does it start?from browsing the site and listing all found commentsfrom the drop-down menu: section, type of pages, language, device, date
Scope of the conclusionfrom a hundred points, among which your reason may or may not beone or two reasons with evidence for each
The role of datethe date usually does not appear at allthe drop start date is checked against updates and your releases
What about irreversiblenot considered: everything looks correctable in the checklistissued in a separate section with an explanation of why it will not be possible to return
Hypothesis testinga new report generation with a new list of itemsrepeated measurements after 4 and 8 weeks using the same method
What we need from you
We can't start without this — best to prepare in advance
1Access to Search Console with history is the main source of diagnostics; without it, we cannot do the work.
2Access to analytics, if available. If it is not there, it will be the first point of the report, not a reproach.
3The history of changes on the site: when the template, engine, structure, texts, hosting were changed - with dates, albeit approximate.
4Access to server logs, if the hosting keeps them, and the answer for how many days they are available.
5One person from your side who remembers what was done on the site half a year ago, or knows who to ask.
If something is missing — let us know, we'll help you gather it or do it as a separate task.
FAQ
Most frequently asked questions — with concrete answers
Where do you start, the site or the data?
From the data. While it's not clear what exactly fell, any check of the site gives a list of general observations unrelated to your fall. We take Search Console, compare two periods by pages, queries, countries and devices and look at the outline: is it the entire site, one section or one language branch. An outline and a date narrow down the cause more than any checklist.
How to distinguish an algorithm update from our own error?
By the shape of the curve and by the contour. The update is usually stretched over several days and hits unevenly across page types. Own error more often gives a cliff in one day and a clear outline: one section, one language, one template. When the date coincides with both the update and your release - and this happens - we separate them by logs and by which addresses stopped receiving impressions.
One language version is down. where to start
From the site map, and this is not a formality. In the cosmetics wholesaler, we counted 4,167 addresses in the canonical map — 3,947 products, 166 brands, 47 categories — and no addresses in a second language version, although bilingualism is implemented and declared via hreflang. In such a situation, the search engine receives a signal that the second version simply does not exist. A fair margin of the mechanism: the map itself does not raise positions, it only makes pages exist for crawling - Google directly documents this.
We have two sitemaps. could this be the reason?
Maybe it happens more often than it seems. Measurement from the same store: the canonical map gives 3,947 products, and the old one lives in parallel at the standard address - 7,782 entries, of which only 2,755 are unique; 2,454 addresses are repeated three times, about 2,706 commercial ones. That is, it is duplicated and incomplete at the same time. We warn you right away: after cleaning duplicates, the number of pages in the index may drop at first. This is an expected course, not a deterioration, but if it is not told in advance, it looks unpleasant.
And if there is no analytics on the site at all?
Then the first point of the audit will be about this. We checked the live pages of a natural cosmetics brand for the EU market and found no counter - no tag manager, no gtag, no pixel. There are 23 products in the catalog, three language versions with correct hreflang, six types of structured data, the first response is 512.6 ms: the site is made neatly, and there is nothing to restore the behavior of visitors. The meter itself does not improve anything, it only gives an opportunity to see - that is why we place the measurement in parallel with the diagnosis, and not after it.
How long to wait after the correction?
Technical stuff is rescanned in 2-6 weeks - hence the 4 and 8 week checkpoints. If there is no movement after eight weeks, the hypothesis was wrong, and we write this directly, instead of suggesting "wait a little longer". Separately: if the reason is that competitors have made better pages, no amount of rescanning will help - there needs to be work on the content, not corrections.
Is it possible to get the traffic back completely?
The honest answer is known only after diagnosis. Technical reasons — the language version has dropped, the redirect is broken, the map returns an empty body — are usually reversible. Loss of items due to stronger competitors can be reversed only with work and definitely not in a week. We include this division in the document as a separate section and do not name the numbers to which we will "return".
Does it happen that the reason is not in the site at all?
Regular. A season, a change in demand, the disappearance of an item from the range, a change in the delivery format at your request - all this causes a drop without any error on your part. Therefore, we compare the dynamics with the demand in the industry before looking for culprits within the site. Otherwise, you can spend a month fixing something that didn't break, and lose the season in that month.
What accesses are needed and what will happen if they are not available?
The key is Search Console with history: it's the only source where you can see impressions and conversions before the drop. Server logs are next in importance: the most accurate way to see a change in bot behavior, but not all hosting stores store them, and the depth is often a week or two. Analytics is desirable, its absence does not block work. Without a webmaster's office, diagnosis turns into guesswork - in this case, we take it up only after access appears.
Submit access to Search Console - we'll do the drop layout ourselves.
In response: what exactly fell, from which day, and what hypothesis of the cause is confirmed by this. If it turns out to be the season, that's exactly what you'll hear.
From measured casesThe canonical map gives 4,167 addresses: 3,947 products, 166 brands, 47 categories, 6 information pages and the main one
There is no price list on the site on purpose: the same work differs several times over between two clients, and a “from” figure explains nothing in that case. First a free audit — we count your pages, duplicates and speed — then we name the sum and the deadline and fix both in the contract.