The Problem

My company had been using a low-cost shared hosting service. I won’t bother to mention who, because many of them are using the same cheap infrastructure.

When I say “slow”, I’m not exaggerating.

If you were lucky enough to load our site at all, it was like trying to download a full-length movie in 1994 with a dial-up modem.

It was that bad.

Beyond the abysmal page load speed, our site was hacked ravaged  several times over the years and that was getting old.

Granted, this wasn’t entirely the fault of our old host. Old plugins and outdated WordPress themes were, in part, to blame. But the old host did fail to proactively defend against these attacks.

The Solution

I’ve used 10+ hosting companies over the past decade. Everything from expensive, pimped out Amazon servers to fly-by-night shared hosting resellers.

I had just finished migrating a different site over to WPEngine, and had an excellent experience with them. In fact, I’m now in the process of migrating all of my sites (including this one). Since we only needed to host a WordPress site, I knew that WPEngine would suit our technical needs (you can check out pricing here)

The Results

We measured results using a number of KPIs. They can be thought of in terms of three different categories:

1. User Engagement: How visitors interact with the site: how much content they consume, how long they hang out, etc.
2. Crawl Stats: The extent to which a bot (namely, Googlebot) is able to discover content throughout the site for indexing in search engine results pages (SERPs).
3. Ranking Changes: Any movement in our Google organic rankings.

Our data originates from three “snapshots” in time: a benchmark (before migration), one week after the migration, and one month after the migration.

The test is as close as we can get to an “all things held equal” scenario. In other words, during this test, we did not actively pursue or earn new external links, didn’t add substantial new content to the site, did not change the site design, etc.

So how did we do?

User Engagement

Only one week later, we saw gains across the board for user engagement metrics. Visitors read through 23% more pages on our site. But we don’t monetize our site through ads, so why care about the extra page views?

It’s a “win” because people are reading more of our content, which extends our influence and helps establish trust. It’s like having a guest over for dinner and they ask for seconds. Feels good, doesn’t it?

Average session duration increased by 62%. Visitors stuck around on the site a lot longer than before. This is like having that same dinner guest never want to leave, presumably, because they’re having a great time.

Bounce rates decreased by 11%. When a visitor “bounces”, they basically entered through one page and never went any further. They were like, “hmm what’s this blog post about. Ok meh, I’m done, let’s bounce“. This is your dinner guest scarfing down a “no thank you” helping and skipping dessert. Not cool. So decreasing that behavior by 11% in the first week was a nice lift to our user engagement (and spirits).

The truth is, none of these KPIs tell the full story by themselves. For instance, more pages per visit can be a bad thing. Maybe the visitor can’t find what they’re looking for, and is frantically clicking through your nav bar. But looking at all of these metrics together tells a more complete story.

One Month Later

Our snapshot one month after the migration to WPEngine showed sustained improvement across all three metrics. Bravo, guys.

Crawl Stats

So, what did Google bot think of the switch to a new host?

Total Pages Crawled

The new hosting plan got Googlebot all hot in bothered (in a good way) immediately after the migration. It peaked at 444 pages crawled across our site just two days later. Compare this to the mere 58 pages that were crawled the day before.

The takeaway here is that something about our new setup told Google that they should take a deeper dive and make sure they haven’t ignored any important content. Truthfully, this high number of pages likely includes low-value stuff like blog categories, tags and other default CMS output that we forgot to “no-index”.

After a month of data collection, the average pages crawled is still roughly 30-40% higher than before, though significantly lower than the peak. That’s cool with us, since Google is finding everything that we really want them to index.

Time Spent Downloading

The time it takes to download one of your pages is a function of your page load speed. Google has gone on record to say that they use page load speed as a factor in their ranking algorithm. That’s the fact. My opinion is that this factor will continue to grow in importance as time marches on.

So this is a big one for us. How did WPEngine do?

Whoa. This is the most excited I’ve been about a graph in awhile.

After the migration we averaged a 71% improvement in page load speed. These gains have been sustained throughout the entire month and we don’t expect that to change.

This graph really catches our hold host with their pants down. Not all hardware is created equally. The WPEngine machines serve our site nearly twice as fast. In fact, the gains we see through Google Webmaster Tools (above) may actually understate the speeds that human users experience, since WPEngine also offers a top-notch caching solution.

Ranking Changes

Great stats so far, but how did this impact our organic rankings in Google?

It’s a tricky question to answer.

For starters, we couldn’t benchmark every keyword that the site previously ranked for and couldn’t possibly discover all of the new rankings that came into the fold after the migration. So the takeaways here are more based on high level observations and patterns, rather than scientific percentage gains/losses.

Takeaways:

1. There was a notable change in ranking movement beyond the first page of results

Of the rankings that changed one month after the benchmark, we saw a disproportionate amount of improvement among the deeper SERPs.

In other words, we saw a lot of improvement of rankings from page 10 search results over to page 3, as opposed to a keyword that was ranking #7 (first page) and moved to #1.

Intuitively, this makes sense. The competition is more formidable on the first page. There is a logarithmic scale to rankings: Small changes have big impact when you’re buried in the SERPs, but it takes more dramatic efforts to crack the first page.

2. Long-tail keyword rankings improved significantly

Most of the ranking improvements we saw were among the long-tail keywords we had rankings for already. These are the niche-specific blog posts and how-to’s.

They also include the keywords with geo-modifiers (the agency is located in Boston).

Long-tail keywords by nature usually have lower competition in the SERPs (not always, but often). So it makes sense that a less potent ranking factor such as page load speed could help move the needle here.

Conclusions:

Site performance goes way beyond the SEO implications. Your website often makes the first impression on customers. Is your site experience laggy and frustrating? Customers may subconsciously perceive that your product or service will deliver a similar experience.

Overall site performance had an impact on our rankings. It primarily impacts long-tail keywords and helps lift pages out of the deep SERPs.

Without changing any content, user engagement metrics improved across the board. Visitors to your site are able to perceive the performance of the “invisible” technology that gives life to the content, so it shouldn’t be an afterthought.

Get Serious About Site Performance

I’m still working on my full review of WPEngine, but so far I’m absolutely thrilled with the results. And that’s just from a technical performance perspective. I haven’t even spoken to the other benefits like site security and the fully-managed platform.

Just as you invest in your content, don’t neglect an investment in your site’s infrastructure.

Disclosure: I’ve recently enrolled as an affiliate for WPEngine. If you use any of my links to check their site, I may receive a commission if you become a customer (at no cost to you). I recommend products that I believe in and know will make a difference – it’s a bonus in this case to have some data that actually supports that belief!