If you ever experienced a major website technology upgrade as an SEO, or currently going through this process, I wrote this case study, which will focus on the process of ensuring your site’s health during such a massive change, the challenges you’ll face, and the solutions I found for a smooth migration.
In my case, the business has decided that one of its big milestones for the year was to upgrade the site’s JS framework from Angular 1.6 to Angular 6 (AKA Angular Universal), while ensuring no regression in organic traffic for the website. And we are talking about a huge site with millions of active users.
Original timelines for this super challenging project were as follow:
Q1 – Q3: JS framework upgrade
Q4: SEO implementation
So, what do you do with this information now? Can we wait with the SEO work while the site works on Angular 6, or need to raise a BIG red flag before development starts? Is the current Dynamic Rendering crawling solution will be supported after the upgrade?
I had no answers at that point, so I decided to take it step by step, and break this complex task to smaller tasks and follow one at a time. Here is what I did:
Step 1: Research
Step 2: Site Audit
Switching to Angular 6
Before switching to Angular 6, I had to understand some key elements that might negatively impact organic performance.
All the visible content that is loaded on the page should not utilize the latest tools that were recently released in Angular 6. For example, if we use tools unique to Angular 6 and newer to load the main navigation, the main body etc – there is a high chance that Google will not be able to load the main navigation or main body. Back then, Google only crawled and rendered using a much older browser (Chrome 41).
Therefore, all the content loaded on the website should still be fully visible, even when you are using an older browser such as Chrome 41.
DOM Content Load Time
How fast the DOM is loaded can have a critical impact on organic performance.
Unlike humans, computers struggle to understand when a page is fully loaded. Generally, most tools depend on network activity. For example, if you run a page speed test using tools such as webpagetest.org, the tool has no idea when the page is actually fully loaded. Therefore, it monitors network activity and determines the page has fully loaded after 2 seconds of no network activity has been detected.
With this in mind, we need to ensure the DOM is continuously loading with no pauses in parsing the DOM. Be it slow server response times from certain API scripts or waiting for an event to fire upon a user-interaction.
Any delays in loading the DOM or dependencies on a user-interaction will be completely missed by Google’s crawler, resulting in a negative impact on organic performance.
A prime example of this can be seen on website homepage, where the main body module takes a bit too long to load. Be it no network activity for a short period of time, causing Google to think the page is finished and ending the render (most likely), or the DOM load time exceeding Google’s timeout.
Therefore, we must ensure any render blocking scripts are removed or moved to the bottom of the page. The DOM is continuously being parsed without any pauses and load times are extremely fast.
Step 3: Data Analysis, Forecast & Potential Risks
Failing to follow the advice above can lead to significant ranking drops, resulting in a loss of traffic and revenue to the business due to any of these issues:
- Loading key content using features only available on Angular 6+
- Loading links using features only available on Angular 6+
- Pauses in the DOM parsing
- DOM full load time taking too long 3+ seconds
In order to understand the impact of ignoring the recommendations above, I have included 3 forecasts to help simulate what could happen. The forecasts were calculated by Avg. Search Volume, Position, and Estimated Traffic (by Avg. CTR per Position model) per ranked keyword positions 1-10.
This forecast presumes that certain new features that are unique to Angular 6 and newer versions are used to load small portion of the website.
This include smaller interactive elements on the page and may include some internal links to deeper event pages but does not include main body text, navigation links etc.
This forecast presumes that certain new features that are unique to Angular 6 and newer versions are used to load several parts of the website, this can include the main body, navigation or other key elements on the site.
This include smaller interactive elements on the page, include all internal links to deeper event pages and in some circumstances main body content.
This forecast presumes that many new features that are unique to Angular 6 and newer versions are used to load a significant parts of the website, this includes the main body, navigation and other key elements on the site.
Step 4: Recommendations
Dev team can utilize Angular 6’s latest features to build all elements of the website’s front-end be it for speed improvements or new capabilities under one condition. If we utilize server-side rendering.
With server-side rendering, we will be passing the parsed DOM directly to users and search engines. This means it does not matter if Google’s crawler does not support Angular 6 as it will not need to execute or attempt to render it.
SEO Checklist for Dev
I wrote down all the required, high level SEO components to ensure developers have a checklist to follow when developing a website in Angular 6, with respect to SEO.Certain features within Angular 6 cannot be interpreted by search engines such as Google. Therefore, incorrectly coding the website can negatively impact organic performance, impacting traffic and revenue.
I provided the dev team a checklist to ensure Google can fully render an Angular 6 built website.
- Using Chrome 41, are all the elements on the page fully visible without any user interaction?
- Angular 6 specific features should only be used to improve the user-experience of a website. For example, after a user has interacted with the website. Due to Google using Chrome 41, we must not use Angular 6 specific features to load any parts of the website, otherwise Google will not be able to render them.
- All links loaded must contain valid HTML hyperlink tags, for example <a href=”URL”>Anchor Text</a>. These links must be visible within the DOM of Chrome 41.
- The website must be built to support server-side rendering soon.
- Ensure the website serves correct status codes. For example, on a nonexistent page, serve 404 status codes and ensure 301 redirects work correctly.
- Pushstate redirects should be avoided throughout the website.
- Main navigation links should be fully visible in the DOM without any user-interaction.
- Tabbed content should be fully visible within the DOM and not dynamically injected upon opening a tab.
- DEV FAQ – Do we still need server-side rendering if we obey all the rules mentioned in this document?
Yes, primarily because the site is very dynamic, and content is always changing, multiple times per day. Although Google can render websites, it does not do it often. This means a newly published event today could take Google a week to see and index it, by that time the event could well have expired.
Server-side rendering will allow Google to see all the content on the website without having depend on rendering it. Resulting in new events/pages being indexed within a matter of minutes/hours.
Step 5: Testing
Now this is where all the fun begins…
Few weeks later, the engineering have completed their first version of the site using Angular 6, on the staging environment, and provided access for testing it for all stake holders.
Apparently, the site couldn’t be crawled by search engines at all, and this is where I realized that there’s no way we can go live with Angular 6 before ensuring organic traffic is not at risk, a true risk.
What to Do When You Hit the Worst-Case Scenario?
It seems that the the impact on organic traffic is going to be unprecedented, if I cannot convince the business not to go live in Q3 with Angular only and without SEO support.
Now this kind of challenge requires more than just good SEO skills. It requires an out of the box thinking and understanding of the internal politics in your organization. If you’re about to severely disrupt the planned project and its time to market, it is not enough to have a solid business case, you also better have a wide internal support from other stake holders and departments to win this battle. You cannot win this alone.
So, I did two things to try and change it:
1. Sharpening my worst-case scenario forecast with extra data and show case it in a more digestible way for the higher management, to try and embrace my recommendations.
This is how I approached it:
- I prepared a list of all keywords ranked positions 1-20, to demonstrate the impact on traffic and revenue in case we lose all rankings due to the framework upgrade. If you want, 1-10 is also enough, I decided to take second page anyway.
- For each keyword I had the following data points:
- Monthly Search Vol.
- Share of Voice (Based on Avg. CTR per position)
- Estimated Traffic (Monthly Search Vol. * Avg. CTR)
- PPC Cost (Estimate Traffic * CPC)
- I calculated the potential monthly SEO revenue loss using the following data points:
- Total Monthly Estimated Traffic (sum Estimated Traffic column)
- Conversion Rate
- User Value $
- Monthly SEO Revenue loss (Total Monthly Estimated Traffic * CVR * User Value)
- PPC Cost to retain traffic (sum PPC Cost column)
|Total Monthly Estimated Traffic||70,000|
|Total Monthly SEO Revenue loss||$142,380|
|Total Monthly PPC Cost to Retain Organic Traffic||$230,099|
Now this kind of data is much easier to understand in terms of bottom line, but why I am using PPC data? For two reasons:
First, you always want to let someone compare between two options, to let him better understand the big picture, and what are the costs associated to each decision he might take.
In my case, I was trying to make a point that retaining lost SEO traffic with PPC is very costly, and has some other side effects, which I will mention later.
Second, to establishing a wide, cross-department shared voice around the preferred solution:
2. Get internal support by using the win-win approach.
I realized that if the business decision will be to wait with the SEO implementation for the 4th quarter, post Angular upgrade, they might think of retaining this traffic with PPC campaigns, which will impact the PPC team as well, in two ways:
- The PPC team will need to recruit more members in order to meet this challenge and all their other goals. Recruiting is a long process, and decisions about our situation need to be done now! And the team is extremely busy anyhow.
- Trying to target long tail search terms with PPC ads is sometimes not doable, and even if it is possible, this will dramatically impact their Quality Score, which might get slashed and potentially risk the entire PPC account.
Once I had the table above and the PPC team on my side, I pitched the business and was able to get their full attention and provide a clear view of what needs to be done.
My Main Takeaways from this challenge
1. SEO education. Product people focus mostly on new features development, that is what they love to do. Dev people do what product tells them to do. No one is really supporting SEO, at least in this scenario. But blame no one other than yourself. We, the SEO people, must educate the business about SEO, if we would like to get their support and understand our needs, but this is a wide topic I tend to address in a separate post.
2. Do not walk alone this path. Leverage internal ‘political’ support to increase your influence on the projected outcomes. Take responsibility.
3. Do not ask for permission, beg for forgiveness. I decided to raise all the red flags in advanced, even if there is a chance I’m wrong in my conclusions. Why? Because I prefer managing the crisis and not being led by it.
Happy to hear your thoughts on this.
You can find the original post in Moshe’s blog