Time to first byte
Just wondering if some of you can list your website, host and what version of TB/PS you’re using and the time to first byte. I tested the demo TB store and it seemed pretty high compared to Prestashop’s demo. Google says you should keep it under 200ms and I know this is gonna become an even more important SEO ranking factor in the coming year. I believe a lot of this has to do with hosting too, so maybe it’s just Cloudways?
TB - http://www.bytecheck.com/results?resource=https%3A%2F%2Ffront.thirtybees.com%2F (I got 917 ms)
420 ms on Google Pagespeed Insights https://developers.google.com/speed/pagespeed/insights/?url=https%3A%2F%2Ffront.thirtybees.com%2F&tab=mobile
Prestashop (wow, really good, unexpected) https://developers.google.com/speed/pagespeed/insights/?url=http%3A%2F%2Fdemo.prestashop.com%2Fen%2F%3Fview%3Dfront&tab=mobile
Is this being looked in to at all by the TB developers? Just curious.
lesley last edited by
It is something we improve with every release actually.
Its a very difficult process to be honest to compare multiple sites with multiple test metrics, but its something I will try to tackle.
The first test you posted, when clicking on the link, this is what I see (I am screenshoting because they vary over time).
thirty bees, http://storage6.static.itmages.com/i/18/0123/h_1516674115_9059172_af10a6ed5d.png
PrestaShop 1.7 http://storage8.static.itmages.com/i/18/0123/h_1516674170_2709716_0682054075.png
I would be lying if I said we beat them, because we clearly arent. But why aren’t we? That is the better question to ask. In a round about way, it is because we care about data security, we care about SEO. To be point blank, it is because we use an SSL certificate and secure the site. Notice that PrestaShop’s demo is non https. Also look in the screen shots I posted and the links you posted. SSL negotiation is taking around 500ms. If we wanted to put out a demo that was pure speed without best practices, we could.
Here is what our demo does with SSL turned off. http://storage2.static.itmages.com/i/18/0123/h_1516674411_7583843_ddcee009c6.png
We beat them in that case.
TTFB is a meaningless metric in my opinion. Like I wrote about in a post last week, what is better a .2 second ttfb and a total page load of 4 seconds? Or a .5ttfb and a full page load of 2 seconds? Are users going to notice the 2. seconds or the 2 seconds? I err on the side of the 2 seconds.
I prefer the raw speed approach. If you look at this test
Which loading time would you prefer? Then scroll down and look at the ttfb. We lose that. Which is more preferable though, a site that loads in .6 seconds or 2 seconds?
In the future its going to get even more complicated that this. Lighthouse is replacing PageSpeed as the preferred metric, its metrics are more meaningful than PageSpeed, it is more of a real world metric. It takes into account not only the loading time of a site, but the processing time as well. I am sure you are like me, you have loaded a site on your phone, it comes up with a blank screen and the bar is spinning. Does that mean the site is still download? It does not actually, your smaller phone processor could actually just be parsing everything that was download.
This is where Lighthouse comes in. It does not just grade on TTFB or complete download of files. It rates off how quick a user can see the first paint of your site and how quickly a user can start using your site. These are more real world and meaningful.
This is something else we win hands down in. In the future it will only get better after we move away from the 1.6 themeing style and more to our own.
Great explanation Leslie. I appreciate it. Should make something like this a sticky post in SEO forum maybe.
Quick question though, if ttfb or ttfp isn’t important in your mind, why does Google show these metrics? Or are you saying Lighthouse is replacing Pagespeed insights and they won’t be checking this metric anymore? And what about “round trips to load render blocking resources”? That seems to be a new metric in insights.
datakick last edited by
Trust me, they know exactly how long it takes for page to show some real content. And since this metric is what matters to people, it definitely matter to ranking algorithm.
lesley last edited by
I have been wondering where I can add all the blog posts I have done about different speed related things. There are quite a few on our blog that break some of these things down, but not one central article. These are some of the better ones
bzndk last edited by bzndk
Your testing methodology is abit off.
If you notice the test you did of TB on Byte Check there is a redirect in the test you did, always make sure you test with the right URL since a redirect can add extra time to most tests, but for me the TTFB of http://front.thirtybees.com/ is about 200-400ms
The way you testet PS is off by a lot, when you test there site you have to remember they use a landing page, where the demo store is loaded in a iframe, this is not countet in to the TTFB, if you go in to the source code and find the iframe, and get the link from there, the results are very different, with the iframe approach the result is 200-300ms – but without the iframe and direct link it hits 400-500ms, so it’s about 200ms slower then TB.
And pagespeed insight is just flaky as hell, a lot of things can be at play here, server location is normally the issue when it says reduce the server reply time, since PageSpeed Insight runs off servers in USA, and I am not sure if PS might use some sort of gateway that sends you to a version of their test site closer to USA when you test with insight.
And again testing sites against each other with different world locations is unfair in general, since some sites will have better scores ex in EU, some will have better in US and so on, the latency has to be taken in to account.
The most important metric is how long it takes before your user can see content and start using the site.
And again theste demo sites never use modules, or have a ton of products, so they will show very different metric compared to a more real world example, the reason i dont mention magento, is that the cost of getting a fully working magento site for production is about 6-8 times higer then Prestashop.