Let me start out with an obvious fact: Site speed matters for SEO.
Any junior SEO could tell you that. But I wanted to know something more important: how and why does page speed matter?
We didn’t realize it when we started this research, but we were about to uncover a specific speed factor that Google uses to rank your website. It’s not just speed alone that matters, but a specific type of page speed that makes the real difference.
The study you’re about to read is the largest study of the impact of website speed on Google search ranking ever performed. Ahrefs contributed ranking data, but it took massive amounts of crawling, collating, and crunching the data on our part to get accurate results.
We surfaced several highly significant findings. These findings are significant, partly because they’ve never been revealed before, but also because they could make a massive impact on your website’s ranking!
Why go to all the trouble to conduct this study?
Some people wonder why I’ve bothered to produce these massive data-driven studies like the one on Hummingbird and Google Local results. Why go to all the time, hassle, and cost to produce this particular study on page speed and Google ranking?
- The last major data-driven study on page speed and Google ranking was done two years ago! We don’t have up to date data analysis on this important topic.
- Google has made massive algorithm changes in the past two years, and has been vocal about the importance of speed.
- User experience and page speed have merged in a way. They influence each other.
- I’m curious! I want to know how my own websites are ranking based on page speed, and what I can do to improve their ranking!
- It boils down to this: I want to give SEOs the most advanced, reliable, and actionable data possible. In the end, it’s foolish to rely on hunches about how SEO works, or how the algorithm operates.
What I want to do is give you rock-solid data that you can bank on 100%. I want you to be able to to take action that gets results.
What we already know about page speed, and what we set out to learn
Like I mentioned already page speed is an important SEO factor. It’s not just important for SEO! A faster website increases conversions, too.
But if you’ve ever dug into pagespeed metrics, you realize that gets way more complicated than just a readout on a clock. In fact, Google’s own PageSpeed Insights doesn’t even give you a timing result!
Helpful as it is, you can’t simply rely on a Pingdom test, for example, to furnish reliable data on page speed, let alone how that speed impacts your Google rankings.
There are so many other factors involved in page speed. What about the impact of plugins, image size, caching, cache validating, server response time, server response codes, time to first byte (TTFB), render rates, server location, DNS lookups, parallelized downloads, encoded headers, static content, code minification, redirects, and all the other factors that play a role?
Like I said, it gets complicated.
I wanted to slice away all the complication, and give you one major insights. (Just one!) But to do so I needed to collect, refine, analyze, parse, configure, and basically digest a whole bunch of data. That’s where those 143,827 URLS come in.
For each URL we measured five following metrics that are part of page speed:
- Time To First Byte: When your browser goes to a URL, it sends a message to the server asking for the HTML document at that URL. This measures the amount of time for the first little bit of website data is sent to your browser. A fast time here might indicate the rest of the site will load quickly.
- Start Render: Rendering happens when a computer turns code into the visual display you see on your screen. Start render is when the first visual part of a website appears. This is useful for users because it let’s them know that the website is doing something.
- Visually Complete: This is when the rendering processed is finished and the site is fully visible to the user.
- Document Complete: Even though the site is visually complete, there are still some operations going on in the background. Once these are finished, the server indicates that the HTML document is finished loading. This is the most technically accurate measure of “site loaded.”
- Fully Loaded: After the document is completed, it is normal for the asynchronous code to start running and loading more items. This does not stop the user from interacting with the site so it is usually not considered to be part of the “load time.” This metric is measured when all loading activity has ceased for 2 seconds.
- Number of file requests: When a site is loaded, it is still requesting multiple files such as CSS, JS, or image files. Loading many small files can unnecessarily slow down loading time, so they should usually be minimized.
Here is how these metrics appear on a speed measurement chart:
For our testing we used the free site speed measurement tool, www.webpagetest.org.
Using an API, we crawled each of the 143,827 URLs to perform extensive speed testing and analysis.
We had a lot of numbers to crunch! It took 100 AWS EC2 servers more than two days to crawl the URLs, run tests on them, and produce metrics.
Here is how we conducted the study:
- Each testing agent was running Chrome on Windows with a 1024×768 desktop screen.
- To determine the URLs to crawl, we generated 5,000 random keywords with a monthly search volume of more than 10 per month in order to ensure we were using a representative sample of all types of keywords, from head terms to longtails.
- We tracked the top 30 Google results for each keyword (non-mobile).
Results of the Study
As expected, we confirmed that faster site speed did correlate with higher Google ranking.
But why? The data surprised us.
First, let’s check out some of the data. Top-ranked websites have high speeds in the “start rendering” metric.
Notice the data in the following chart. The numbers on the x-axis (at the bottom of the chart) indicate the website’s Google rank position. The y-axis denotes speed in milliseconds.
The chart indicates that overall load time is especially faster for the first five positions. Rank 6 was, on average, 20% slower than rank 1.
The data suggests that improving start render times contributes directly to a higher Google result.
Additionally, we corroborated the findings of other research, that top-ranked websites have faster speeds for time to first byte (TTFB). This was the largest correlation noted in our study. Notice how the chart below, tracking only TTFB speeds, has a massive increase in speed between positions 1, 2, and 3.
Notice how TTFB correlation (chart above) is far higher than start render (chart below).
To make the point visually again, TTFB has significant differentiation from the next closest metric, which is start render (chart below).
Averaged out across all 143,827 URLs, here is what we found:
Now, take a look at the correlation ranking of each of these page speed measurements:
Because TTFB has such a pronounced impact on ranking, it’s easy to overlook the impact of of the lesser significant speed rankings.
Although TTFB correlation is high, our research uncovered a new angle to site speed and ranking impact. Google does not consider a simplified TTFB score alone as they did in the past. Today, they are looking at the more sophisticated interplay of doc complete and start render, and using these speed factors as ranking signals in addition to the TTFB measurements.
This angle of observation goes beyond the simple “fix your TTFB problems and all will be well.” As our research uncovered, Google is doing a far more accurate job of analyzing meaningful speed metrics.
It is strategic to address doc complete and start render timing metrics, because these can accrue to improved rankings.
Why the big emphasis on TTFB?
What’s going on here? Why is TTFB such a massively significant ranking factor?
I tend to think it boils down to user experience. Have you ever read Google’s company philosophy? They wrote these ten principles when the company was young.
- The most important truth is this: Focus on the user and all else will follow.
- The third most important truth is this: Fast is better than slow.
Taken together, it seems a safe assumption that TTFB is a significant user factor. If Google gives a site with high TTFB speed a higher ranking, this ranking uptick suggests that they perceive this site as having a superior user experience.
That’s why our research is significant. By plotting the correlative significance of “lesser” site speed metrics — document complete and start render in particular — we can see that their cumulative impact is nearly as great as TTFB’s.
How to improve your TTFB
Most basic SEO articles will tell you how to improve your site speed. However, this is not a basic SEO article, nor do you need to hear someone say “improve your site speed.” Instead, you need to focus on improving specific aspects of a page speed.
You can start by measuring your site’s current TTFB.
First, go to www.webpagetest.org, and enter your website’s URL. Click “Start Test.”
Adjust the “test location” to find the server that is geographically closest to where the majority of your users are. If you know that your users use a browser other than Chrome, you can also change the browser setting.
I recommend keeping the “Advanced Settings” as they are unless you have some experience or advanced knowledge about speed tests that would influence the results you’re looking for.
The test usually takes less than one minute to complete. When it is finished, look for the “first byte” number in the chart.
This is your page’s TTFB. If you’re really geeking out on the data at this point, click the “first view” chart for an expanded view of the waterfall, connection, and request details.
The “request details” chart tabulates the TTFB for all the site’s resources:
How do you know if your TTFB needs improving? Here is a suggested guide:
Do you think your TTFB needs improvement? These six methods are guaranteed to amp up your TTFB. (Pro tip: Hand this list to your developer. He or she will know what to do.)
- Use a CDN.
- Optimize application code.
- Optimize database queries.
- Reduce HTTP requests.
- Ensure fast server response time.
- Use a respond first, process later (RFPL) method for caching.
Conclusion and three really important takeaways
Takeaway #1: Speed is important, but it’s just one ranking factor among many.
Like most ranking factors, page speed alone won’t earn you a top spot on the SERPs, but a slow speed will hold you back from ranking higher. So what should you do? Try to improve your site speed, but keep optimizing your other search factors as well.
Takeaway #2: UX is super important.
Keep in mind as well that Google really likes a good UX. Stripping down your site to a few bare-bones lines of HTML won’t necessarily boost your rankings, however, if you’ve destroyed the visual appeal or overall UX of the site in the process. You need both to optimize speed and user experience to improve your rankings.
Takeaway #3: Optimize your TTFB first!
When you’re in the weeds of optimizing your site speed, play smart. Instead of trying hard to push down your document complete, start rendering, or fully loaded page speed metrics (all of which are important), focus on the highest-impact metric first: TTFB.
What’s next? Call your developer, or roll up your sleeves and start tackling TTFB! You’ll be super glad you did!
I’m always eager to hear success stories — how data like this has a real world impact. Tell me what changes you made on your site, and how it improved things!