The proportion of the search results page that has been contaminated exceeds eight percent, what should the content station traffic analysis exclude first?

The ratio of polluted search result pages exceeds 8%, what should be excluded in the analysis of content station traffic? First, let's talk about the meaning of these two terms: In the third-party crawling, a certain proportion of daily returned search results show degradation, and the proportion of pollution identified during this period has increased significantly, while the baseline before was only 0.17 percentage points. The direct inference of this observation is that the decline in your natural traffic data may not come from changes in content quality. Before analysis, it is necessary to exclude three types of noise: abnormal traffic on the crawling side, coverage gaps on the collection side, and distortion in word-page matching.

What does it mean by polluted search result pages

The search results degradation is usually not due to the algorithm '降低你', but rather the result page itself being occupied by things that should not be there: pages copied in bulk by collection stations, jump pages stuffed with keywords, and garbage content hung on by intruded sites.

There are two paths of actual impact on the content site. One is the click being intercepted, where users click on the content page ranked in front but not relevant; the other is data distortion, where non-target traffic visits are mixed into the entry path you see in the analysis backend, diluting the conversion indicators.

Determine if it is a problem of this type, check if the competitor's page under the same keyword also shows abnormality in the same time period, as well as whether a large number of irrelevant matching words pop up in your ranking records.

The first category to be excluded: crawling side anomaly

Robot traffic is the easiest to disguise as 'algorithmic impact'. The characteristics are that the traffic volume increases but business actions do not increase, the bounce rate and dwell time are significantly deviated from the normal, and the requests are concentrated on a few paths.

Here the robots configuration is responsible for range control: which directories are allowed to be crawled and which should not enter the search view need to be clear. What needs to be blocked is the interference of low-value crawling on quotas and logs, and this part should be separated from the analysis data before discussing the trend.

The second category excluded: the gap in collection

The second type of noise is that the data seems to have dropped, but in fact the page has never entered the candidate set.

First, check the site map: after column adjustments or batch imports, confirm that the new page is within the coverage range of the sitemap and that the generation task has indeed run. Then, check the consistency of the address: historical addresses have not been redirected, the weight remains on the old address, and the new address is not stable in the results. Finally, check the interface reading: the high-frequency capabilities such as document list and details should be consistent with the content returned by the API call, and the interface should not return a batch of pages, or the crawling side will also find fewer pages.

These three items are not reflected in the analysis background, but they will make the 'traffic drop' look inexplicable.

The third category to be excluded: word and page matching distortion

The third category is attribution error. The keywords in the report often mismatch the real needs: brand words mixed with general words, long-tail queries truncated, the same intent scattered across multiple synonyms.

At this point, align with two data sources. One is the keyword library, which manages words with business value, corresponds to each word to the landing page, so that the report has a comparable object. The other is the in-site search words, which are the needs written by the users themselves, more direct than the display words in the search console, especially those queries that appear in the in-site search but do not bring clicks - these are the gaps in content coverage.

After aligning two types of data, the remaining decline is worth treating as a content issue.

When is it really time to move content?

Move when three conditions are met simultaneously: noise has been excluded, decline is concentrated on words with commercial intent, and there is a noticeable matching deviation on the corresponding page. If not met, the more appropriate action is to adjust the configuration and supplement the coverage, rather than rewriting the text.

Common questions

During the rising period of pollution ratio, should we continue to publish new content?

Continue. Stopping the press conference will further shrink the coverage of inclusion, rather than magnifying the range of fluctuations.

Which is more可信 between the search terms within the site and the words in the data background?

The purpose is different. In-site search terms reflect real needs, suitable for finding content gaps; back-end display terms reflect external entry paths, suitable for observing trends. They should be read separately.

How to confirm that the garbage traffic has been removed

Group abnormal visits separately by path and UA and observe them for a period of time to see if business actions are still concentrated in the remaining traffic. After grouping, if the indicators return to normal, it means that the interference has been separated.