How to Tell If a Slow Website Is Your Host or Your Own Pages
Your site crawls and your host says everything is fine. Here is how to prove whether the delay is the server or your own pages, in ten minutes.
Your website feels slow, your host's support page says everything is fine, and you are stuck in the middle wondering who to believe. The argument goes nowhere until you have a number that splits the blame, and one measurement does that. You can run it in about ten minutes with free tools.
The one number that splits the blame
Every page load has two halves. The first half is the wait for the server to start answering, called time to first byte, or TTFB. The second half is everything after that: downloading the HTML, images, stylesheets and scripts, then rendering them. If the first half eats most of the time, the delay lives on the server side: your host, your application, or your database. If the second half eats most of it, the delay lives in your own pages, and a faster hosting plan will not fix it.
Web.dev, Google's own guidance, calls a TTFB of 0.8 seconds or less good, and anything over 1.8 seconds poor. Those are rules of thumb, not laws. A page that takes 4 seconds to load with a TTFB of 0.3 seconds is a page problem. A page that takes 4 seconds with a TTFB of 3 seconds is a server problem. Same total, opposite fix.
Step 1: Get a measurement from outside your own connection
Start by ruling out your own network. Load the slow page from a different connection and a private window with extensions switched off. If it is fast everywhere except your own browser, you are looking at a local problem, not a site problem, and you can stop here.
Then run the page through our Website Speed Test, which measures it from outside your connection. Note the two numbers it gives you: the time to first byte, and the total load time. The gap between them is the story of this post.
Step 2: Break the request into stages with curl
If you have a terminal, one curl command splits the wait into its stages:
curl -s -o /dev/null -w "dns: %{time_namelookup} connect: %{time_connect} tls: %{time_appconnect} ttfb: %{time_starttransfer} total: %{time_total} size: %{size_download}" https://yoursite.com/
Read it like a triage nurse. DNS high, over a tenth of a second repeatedly, points at a DNS problem. Connect or TLS high points at network distance, which a CDN usually erases. TTFB minus TLS high means the server is thinking, and that is your application or database. Total minus TTFB high means payload weight, a front-end problem. Run it three times and trust the pattern.
A worked example from this very site
I ran that command on websmartools.com this morning, and the numbers tell the story better than any theory. The homepage answered with a TTFB of 0.75 seconds and a total of 0.81 seconds, at 14,815 bytes. The static robots.txt file on the same host answered with a TTFB of 0.44 seconds and a total of 0.48 seconds, at 153 bytes.
Read those two lines together and you can see the split. The static file answers about 0.3 seconds faster than the homepage on the same server, which is the cost of the application building the page instead of handing over a file. The total minus TTFB gap is tiny, 0.06 seconds on the homepage. If the homepage had shown a TTFB of 0.4 seconds and a total of 3 seconds, the diagnosis would be the opposite: a healthy server serving a heavy page. The same numbers, run on your site, give you the same verdict.
The checks that point at your own pages
Three signs say the problem is yours, not your host's. First, a healthy TTFB with a long total, as above. Second, a big download size: a page shipping several megabytes of images, fonts and scripts will be slow on any host, and the fix is compression, resizing and removing what you do not need. Third, two of Google's Core Web Vitals, which the host cannot move at all. Interaction to Next Paint, good at 200 milliseconds or less, is almost entirely main-thread JavaScript. Cumulative Layout Shift, good at 0.1 or less, comes from images without dimensions, late banners and web fonts. If either of those is poor while TTFB is fine, stop shopping for hosts and start trimming the page. Our post on tracing why a page takes so long to start loading walks through the page side in more detail.
The checks that point at your host
Three signs point the other way. First, a TTFB that is high on every page, including a static file like robots.txt. If even a 153-byte file takes over a second to start arriving, the server itself is struggling, and no amount of image compression will help. Second, slowness that comes and goes. Test the same page at a quiet hour and again at peak traffic: a big gap between the two points at server load, a crowded shared server, or a cache that expires and gets rebuilt. Constant slowness points at page weight or a slow application; intermittent slowness points at capacity. Third, a page that is slow from every location and every network. That comparison is the one worth sending to support, because it turns "my site is slow" into "this server is slow", which is a claim they can act on.
The mistake I see most often
The most common mistake is upgrading the hosting plan before measuring anything. I understand the impulse: the host is the one thing you pay for, so it feels like the lever. But if your TTFB is 0.4 seconds and your page ships 4 megabytes, a bigger server changes nothing. The mirror-image mistake is blaming the host for a page that loads a hero image at full camera resolution, then wondering why the new host did not help either. Measure first, spend second.
What to do first
Run the Website Speed Test on your slowest page and write down the TTFB and the total. If TTFB is most of the total, run the curl breakdown, compare a static file against a dynamic page, and test again at a quiet hour. Then take those numbers to your host, or set up our Uptime Monitor to watch the response time over a week so you have a record, not a feeling. If the total dwarfs the TTFB, start with the heaviest files on the page and work through the checks in our post on fixing a slow website, in order. Either way, you now know which half of the load to fix, and that settles the argument in ten minutes.
A new post lands here every day. If there is a topic you want covered, email ask@websmartstudios.com.