Contextual investigation: How the Cookie Monster Ate 22% of Our Visibility



Last year, the group at Homeday — one of the main property tech organizations in Germany — settled on the choice to relocate to another substance the executives framework (CMS). The objectives of the movement were, in addition to other things, sped up and making a best in class, future-evidence site with every one of the important elements. One of the primary inspirations for the movement was to empower content editors to work all the more unreservedly in making pages without the assistance of designers.


Subsequent to assessing a few CMS choices, we settled on Contentful for its advanced innovation stack, with a prevalent encounter for the two editors and engineers. From a specialized perspective, Contentful, as a headless CMS, permits us to pick which delivering procedure we need to utilize.


We’re as of now completing the movement in a few phases, or waves, to diminish the gamble of issues that have an enormous scope adverse consequence. During the primary wave, we experienced an issue with our treat assent, which prompted a perceivability loss of practically 22% in five days or less. In this article I’ll portray the issues we were looking during this first movement wave and how we settled them.


Setting up the primary test-wave

For the main test-wave we picked 10 SEO pages with high traffic yet low change rates. We laid out a framework for detailing and checking those 10 pages:


Rank-following for most pertinent watchwords


Website design enhancement dashboard (DataStudio, Moz Pro, SEMRush, Search Console, Google Analytics)


Standard slithers


After a complete preparation and testing stage, we relocated the initial 10 SEO pages to the new CMS in December 2021. Albeit a few difficulties happened during the testing stage (expanded stacking times, greater HTML Document Object Model, and so forth) we chose to go live as we didn’t see enormous blocker and we needed to relocate the first testwave before christmas.


First execution survey

Exceptionally amped up for accomplishing the initial step of the movement, we investigated the exhibition of the relocated pages on the following day.


What we saw next truly didn’t satisfy us.


Short-term, the perceivability of followed catchphrases for the relocated pages diminished from 62.35% to 53.59% — we lost 8.76% of perceivability in one day!


Because of this precarious drop in rankings, we led another broad round of testing. In addition to other things we tried for inclusion/ordering issues, if all meta labels were incorporated, organized information, inner connections, page speed and versatility.


Second execution audit

Every one of the articles had a store date after the movement and the substance was completely recorded and being perused by Google. Additionally, we could reject a few movement risk factors (change of URLs, content, meta labels, design, and so on) as wellsprings of blunder, as there hasn’t been any changes.


Perceivability of our followed catchphrases experienced one more drop to 40.60% throughout the following couple of days, making it a complete drop of practically 22% in five days or less. This was additionally obviously displayed in contrast with the opposition of the followed catchphrases (here “assessed traffic”), yet the perceivability looked similar to.


Information from SEMRush, indicated watchword set for followed catchphrases of relocated pages

As other relocation risk factors in addition to Google refreshes had been barred as wellsprings of blunders, it most certainly must be a specialized issue. An excessive amount of JavaScript, low Core Web Vitals scores, or a bigger, more perplexing Document Object Model (DOM) could be in every way likely causes. The DOM addresses a page as items and hubs so that programming dialects like JavaScript can associate with the page and change for instance style, design and content.


Following the treat scraps

We needed to distinguish issues as fast as could really be expected and do speedy bug-fixing and limit more adverse consequences and traffic drops. We at long last got the main genuine sprinkle of which specialized reason could be the reason when one of our devices showed us that the quantity of pages with high outside connecting, as well as the quantity of pages with most extreme substance size, went up. Pages genuinely must don’t surpass the most extreme substance size as pages with an exceptionally enormous measure of body content may not be completely filed. With respect to high outside connecting it is vital that all outer connections are dependable and pertinent for clients. It was dubious that the quantity of outside joins went up very much like this.


Increment of URLs with high outside connecting (more than 10)

Increment of URLs which surpass the predetermined greatest substance size (51.200 bytes)


The two measurements were excessively high contrasted with the quantity of pages we relocated. Be that as it may, why?


In the wake of checking which outside joins had been added to the moved pages, we saw that Google was perusing and ordering the treat assent structure for all relocated pages. We played out a site search, checking for the substance of the treat assent, and saw our hypothesis affirmed:


A site search affirmed that the treat assent was recorded by Google

This prompted a few issues:


There was lots of copied content made for each page because of ordering the treat assent structure.


The substance size of the relocated pages radically expanded. This is an issue as pages with an exceptionally huge measure of body content may not be completely filed.


The quantity of outer active connections definitely expanded.


Our scraps unexpectedly showed a date on the SERPs. This would propose a blog or news story, while most articles on Homeday are evergreen substance. Also, because of the date showing up, the meta depiction was cut off.


However, for what reason was this occurrence? As per our specialist organization, Cookiebot, web search tool crawlers access sites reproducing a full assent. Thus, they get close enough to all satisfied and duplicate from the treat assent flags are not recorded by the crawler.


So for what reason wasn’t this the situation for the relocated pages? We slithered and delivered the pages with various client specialists, yet at the same time couldn’t find a hint of the Cookiebot in the source code.


Examining Google DOMs and looking for an answer

The moved pages are delivered with dynamic information that comes from Contentful and modules. The modules contain just JavaScript code, and now and again they come from an accomplice. One of these modules was the treat director accomplice, which gets the treat assent HTML from outside our code base. For that reason we didn’t find a hint of the treat assent HTML code in the HTML source documents in any case. We saw a bigger DOM yet followed that back to Nuxt’s default, more intricate, bigger DOM. Nuxt is a JavaScript system that we work with.


To approve that Google was perusing the duplicate from the treat assent pennant, we utilized the URL review instrument of Google Search Console. We analyzed the DOM of a relocated page with the DOM of a non-moved page. Inside the DOM of a moved page, we at long last found the treat assent content:


Inside the DOM of a moved page we found the treat assent content

Something different that stood out were the JavaScript documents stacked on our old pages versus the records stacked on our relocated pages. Our site has two contents for the treat assent pennant, given by an outsider: one to show the standard and snatch the assent (uc) and one that imports the flag content (album).


The main content stacked on our old pages was uc.js, which is liable for the treat assent pennant. It is the one content we really want in each page to deal with client assent. It shows the treat assent standard without ordering the substance and recoveries the client’s choice (assuming they concur or differ to the use of treats).


For the moved pages, beside uc.js, there was additionally a cd.js document stacking. Assuming that we have a page, where we need to show more data about our treats to the client and record the treat information, then, at that point, we need to utilize the cd.js. We felt that the two documents are reliant upon one another, which isn’t right. The uc.js can run alone. The cd.js record was the motivation behind why the substance of the treat pennant got delivered and ordered.


It required a long time to find it since we thought the subsequent document was only a pre-necessity for the first. We established that essentially eliminating the stacked cd.js record would be the arrangement.


Execution survey in the wake of carrying out the arrangement

The day we erased the document, our catchphrase perceivability was at 41.70%, which was as yet 21% lower than pre-movement.


Be that as it may, the day subsequent to erasing the document, our perceivability expanded to 50.77%, and the following day it was practically back to ordinary at 60.11%. The assessed traffic acted in much the same way. What a consolation!


Rapidly in the wake of executing the arrangement, the natural traffic returned to pre-movement levels


I can envision that numerous SEOs have managed minuscule issues like this. It appears to be minor, yet prompted a huge drop in perceivability and traffic during the movement. Therefore I propose moving in waves and hindering sufficient time for exploring specialized mistakes when the movement. Also, keeping a nearby gander at the site’s exhibition soon after the movement is pivotal. These are certainly my vital important points from this relocation wave. We just finished the subsequent relocation wave in the start of May 2022 and I can express that up to this point no significant bugs showed up. We’ll have two additional waves and complete the relocation ideally effectively toward the finish of June 2022.


Next Post