Case Study: How the Cookie Monster Ate 22% of Our Visibility

Final 12 months, the staff at Homeday — one of many main property tech corporations in Germany — made the choice emigrate to a brand new content material administration system (CMS). The targets of the migration have been, amongst different issues, elevated web page velocity and making a state-of-the-art, future-proof web site with all the required options. One of many major motivators for the migration was to allow content material editors to work extra freely in creating pages with out the assistance of builders. 

After evaluating a number of CMS choices, we selected Contentful for its trendy know-how stack, with a superior expertise for each editors and builders. From a technical viewpoint, Contentful, as a headless CMS, permits us to decide on which rendering technique we wish to use. 

We’re at present finishing up the migration in a number of phases, or waves, to scale back the chance of issues which have a large-scale adverse influence. Through the first wave, we encountered a problem with our cookie consent, which led to a visibility lack of nearly 22% inside 5 days. On this article I will describe the issues we have been dealing with throughout this primary migration wave and the way we resolved them.

Organising the primary test-wave 

For the primary test-wave we selected 10 website positioning pages with excessive visitors however low conversion charges. We established an infrastructure for reporting and monitoring these 10 pages: 

• Rank-tracking for many related key phrases 

• website positioning dashboard (DataStudio, Moz Professional,  SEMRush, Search Console, Google Analytics)

• Common crawls 

After a complete planning and testing phase, we migrated the primary 10 website positioning pages to the brand new CMS in December 2021. Though a number of challenges occurred in the course of the testing section (elevated loading instances, larger HTML Doc Object Mannequin, and many others.) we determined to go reside as we did not see huge blocker and we needed emigrate the primary testwave earlier than christmas. 

First efficiency evaluation

Very enthusiastic about attaining step one of the migration, we took a take a look at the efficiency of the migrated pages on the subsequent day. 

What we noticed subsequent actually did not please us. 

In a single day, the visibility of tracked key phrases for the migrated pages diminished from 62.35% to 53.59% — we misplaced 8.76% of visibility in someday! 

Because of this steep drop in rankings, we carried out one other intensive spherical of testing. Amongst different issues we examined for protection/ indexing points, if all meta tags have been included, structured information, inside hyperlinks, web page velocity and cell friendliness.

Second efficiency evaluation

All of the articles had a cache date after the migration and the content material was absolutely listed and being learn by Google. Furthermore, we might exclude a number of migration danger components (change of URLs, content material, meta tags, format, and many others.) as sources of error, as there hasn’t been any modifications.

Visibility of our tracked key phrases suffered one other drop to 40.60% over the subsequent few days, making it a complete drop of virtually 22% inside 5 days. This was additionally clearly proven compared to the competitors of the tracked key phrases (right here “estimated visitors”), however the visibility seemed analogous. 

As different migration danger components plus Google updates had been excluded as sources of errors, it positively needed to be a technical challenge. An excessive amount of JavaScript, low Core Net Vitals scores, or a bigger, extra advanced Doc Object Mannequin (DOM) might all be potential causes. The DOM represents a web page as objects and nodes in order that programming languages like JavaScript can work together with the web page and alter for instance fashion, construction and content material.

Following the cookie crumbs

We needed to determine points as shortly as doable and do fast bug-fixing and decrease extra adverse results and visitors drops. We lastly received the primary actual trace of which technical purpose may very well be the trigger when considered one of our instruments confirmed us that the variety of pages with excessive exterior linking, in addition to the variety of pages with most content material dimension, went up. It is vital that pages do not exceed the utmost content material dimension as pages with a really great amount of physique content material is probably not absolutely listed. Relating to the excessive exterior linking it can be crucial that every one exterior hyperlinks are reliable and related for customers. It was suspicious that the variety of exterior hyperlinks went up identical to this.

Each metrics have been disproportionately excessive in comparison with the variety of pages we migrated. However why?

After checking which exterior hyperlinks had been added to the migrated pages, we noticed that Google was studying and indexing the cookie consent type for all migrated pages. We carried out a website search, checking for the content material of the cookie consent, and noticed our concept confirmed: 

This led to a number of issues: 

1. There was tons of duplicated content material created for every web page as a result of indexing the cookie consent type. 

2. The content material dimension of the migrated pages drastically elevated. This can be a downside as pages with a really great amount of physique content material is probably not absolutely listed. 

3. The variety of exterior outgoing hyperlinks drastically elevated. 

4. Our snippets all of a sudden confirmed a date on the SERPs. This might recommend a weblog or information article, whereas most articles on Homeday are evergreen content material. As well as, as a result of date showing, the meta description was lower off. 

However why was this occurring? In response to our service supplier, Cookiebot, search engine crawlers access websites simulating a full consent. Therefore, they acquire entry to all content material and copy from the cookie consent banners usually are not listed by the crawler. 

So why wasn’t this the case for the migrated pages? We crawled and rendered the pages with totally different consumer brokers, however nonetheless could not discover a hint of the Cookiebot within the supply code. 

Investigating Google DOMs and trying to find an answer

The migrated pages are rendered with dynamic information that comes from Contentful and plugins. The plugins comprise simply JavaScript code, and generally they arrive from a companion. Considered one of these plugins was the cookie supervisor companion, which fetches the cookie consent HTML from outdoors our code base. That’s the reason we did not discover a hint of the cookie consent HTML code within the HTML supply information within the first place. We did see a bigger DOM however traced that again to Nuxt’s default, extra advanced, bigger DOM. Nuxt is a JavaScript framework that we work with.

To validate that Google was studying the copy from the cookie consent banner, we used the URL inspection device of Google Search Console. We in contrast the DOM of a migrated web page with the DOM of a non-migrated web page. Throughout the DOM of a migrated web page, we lastly discovered the cookie consent content material:

One thing else that received our consideration have been the JavaScript information loaded on our previous pages versus the information loaded on our migrated pages. Our web site has two scripts for the cookie consent banner, supplied by a third occasion: one to point out the banner and seize the consent (uc) and one which imports the banner content material (cd).

• The one script loaded on our previous pages was uc.js, which is answerable for the cookie consent banner. It’s the one script we’d like in each web page to deal with consumer consent. It shows the cookie consent banner with out indexing the content material and saves the consumer’s choice (in the event that they agree or disagree to the utilization of cookies).

• For the migrated pages, apart from uc.js, there was additionally a cd.js file loading. If we’ve got a web page, the place we wish to present extra details about our cookies to the consumer and index the cookie information, then we’ve got to make use of the cd.js. We thought that each information are depending on one another, which isn’t right. The uc.js can run alone. The cd.js file was the rationale why the content material of the cookie banner received rendered and listed.

It took some time to search out it as a result of we thought the second file was only a pre-requirement for the primary one. We decided that merely eradicating the loaded cd.js file could be the answer.

Efficiency evaluation after implementing the answer

The day we deleted the file, our key phrase visibility was at 41.70%, which was nonetheless 21% decrease than pre-migration. 

Nevertheless, the day after deleting the file, our visibility elevated to 50.77%, and the subsequent day it was nearly again to regular at 60.11%. The estimated visitors behaved equally. What a reduction! 

Conclusion

I can think about that many SEOs have handled tiny points like this. It appears trivial, however led to a big drop in visibility and visitors in the course of the migration. This is the reason I recommend migrating in waves and blocking sufficient time for investigating technical errors earlier than and after the migration. Furthermore, retaining a detailed take a look at the location’s efficiency throughout the weeks after the migration is essential. These are positively my key takeaways from this migration wave. We simply accomplished the second migration wave at first of Might 2022 and I can state that up to now no main bugs appeared. We’ll have two extra waves and full the migration hopefully efficiently by the tip of June 2022.

The efficiency of the migrated pages is nearly again to regular now, and we’ll proceed with the subsequent wave.