Analyzing someone else’s project almost always starts the same way: traffic is falling, several contractors have worked on the site before you, and there’s no clear record of changes. Below is a process that saves you the first few days of work, along with practical advice on duplicates and cannibalization: where pages really compete with each other, and where the concern is overblown.
Start the analysis with data exports, not a technical audit
It’s tempting to open a crawler and start looking for broken links, but on a declining project, that rarely gives you an answer. First, you need data on how demand and visibility have changed.
A practical sequence:
- export the entire available period from Search Console—impressions, clicks, CTR, query list, and pages;
- do the same with Yandex Webmaster;
- combine both exports in one table and compare trends for each search engine separately.
The point of the comparison is that trends across the two search engines answer the key diagnostic question. If traffic has fallen at the same time in Yandex and Google, the cause is most likely on the site itself or in overall demand: seasonality, a site migration, structural changes, or lost pages. If one system has dropped while the other holds steady, that points to an algorithm update, a penalty, or a specific filter—and you should focus your investigation there.
Until you’ve exported the data, it’s impossible to tell an algorithmic demotion from a seasonal decline. The hypothesis “looks like a filter” without an impressions graph is just guesswork, and on a project with five previous contractors, it’s almost certain to send you in the wrong direction.
It’s also worth checking which pages used to get clicks and looking at their current rankings. Often, it turns out that it’s not the whole site that has declined, but a particular section, which narrows the scope of the work.
Duplicate titles: what happens in practice
A classic case: a blog article about “galvanizing” and a commercial service page for “galvanizing.” The titles are almost identical, and both pages are targeting the same query.
In theory, with mixed intent in the SERPs, both could make the top results—both the informational article and the service page. This does happen in practice, but as an exception, not a rule. According to specialists who have analyzed hundreds of thousands of duplicates, the typical scenario is different: the pages compete with each other, the search engine can’t pick an obviously more relevant one, and neither wins. When planning site structure, don’t count on both pages holding their positions in the top results—it’s a lottery, not a strategy.
A separate note on Yandex. One approach is to track cannibalizing queries hourly and check for overlapping URLs in the SERPs. These measurements show that two of your own pages don’t appear in Yandex’s results for the same query—there’s no overlap at all. In other words, the hope that “we’ll leave both up and one will catch on” doesn’t work for Yandex: there’s still only one spot.
How to check pages for cannibalization
The issue arises when there is a category and several clarifying subcategories based on a single property: “spindles,” “spindles with a cross-section of 20,” “spindles with a cross-section of 50.” The question is how many pages to create.
SERP-based clustering gives you a reliable answer, rather than comparing the texts themselves. Look at what the search engine actually shows for each query, then count overlapping URLs in the top results:
- if the keywords fall into different clusters, create separate pages;
- if they fall into the same cluster, build one page and don’t create unnecessary pages.
A common mistake is to decide based on content volume: “There isn’t enough text for three pages, so I’ll combine them into one.” That’s risky logic. If the SERPs show three different clusters, after merging, only one page will rank—the one you kept as the primary page—and you’ll simply lose the demand for the other two. Content volume is secondary here: what justifies a page is different intent, not the number of characters.
The rule works the same way for Yandex and Google, so it’s convenient to cluster once for both systems and compare any differences.
Where to assign a compound keyword
Another common case: your keyword list includes “cafe automation,” “restaurant automation,” and “cafe and restaurant automation.” The first two clearly belong on separate pages, while the third falls between them.
There’s no point in creating a separate page for it: it isn’t a new intent, just a combination of two existing ones. A more practical approach is to check the SERPs for the compound query and assign the keyword to the page with more URL matches in the top results. The logic is simple: a page that already ranks can take on more keywords, while a new page starts from scratch and is very likely to become a third competitor to your own sections.
A tiny issue that can tank rankings
There’s a class of problem that neither an audit nor analytics will catch: a Latin letter inside a Cyrillic word. One substituted “с” in a title, and rankings for the query drop, even though the site looks completely fine.
This can happen in different ways: switching keyboard layouts, copying from an external document, or editing through the admin panel on someone else’s device. It can also be found by chance—for example, when you paste a copied title into the search box and Yandex highlights a typo in a word that looks correctly spelled.
It’s worth checking for this deliberately: run titles and H1 through a search, keep a browser extension that highlights mixed alphabets, or add a regex for mixed characters to your template-checking script. On a large catalog, one discovery is enough to make this check worthwhile.
When you have one working day for a project
Prioritization is a separate question: you can’t catch up with competitors on high-volume keywords, the informational section brings traffic but no leads, and you have just one day.
In a day, you can select low-volume queries with manageable competition and optimize a dozen pages for them. It’s not the strongest move over the course of a year, but it’s the only one that fits the time available and delivers a measurable result.
Strategically, new sections and new projects really are the main growth opportunity: you can improve existing pages, but the payoff takes time. But you can’t build a review site, catalog, or roundup in just one day. Even with AI generation, you need to collect data first, and there’s nowhere to get reviews—generated reviews won’t solve the problem.
Workflow
- Export Search Console and Webmaster data for the entire available period, then compare trends across the systems.
- Determine whether the whole site has declined or only specific sections, and narrow the scope of your analysis.
- Make a list of queries where two of your pages show up in the SERPs—these are potential cannibalization cases.
- Run disputed groups through SERP-based clustering and decide what to do with pages based on the clusters, not the amount of text.
- Assign compound keywords to the page that already ranks.
- Check titles for mixed alphabets.
- If you’re short on time, focus on low-volume keywords and optimize a limited list of pages instead of starting a structural overhaul.
