This document summarizes performance metrics for WordPress websites from the HTTP Archive and Chrome UX Report. Key findings include: the median page weight for WordPress sites is 1.9 MB with significant growth in CSS and image weights; first contentful paint, DOM content loaded, and onload times skew slower for WordPress sites compared to averages; and monitoring performance metrics can help the WordPress ecosystem improve performance over time through best practices, case studies, and incentives for developers.
6. 6wpostats.comWPO Stats: Case studies correlating performance and conversions
“Rebuilding Pinterest pages for performance
resulted in a 40% decrease in wait time, a
15% increase in SEO traffic and a 15%
increase in conversion rate to signup.”
9. 9bit.ly/2HjmjdUMalte Ubl, TL of AMP
"Objective metrics
collected through
public data sets like the
Chrome User
Experience Report
can provide for
objective performance
and UX data
independent of
technology."
10. 10Source: “The New Bar for Web Experiences” at Chrome Dev Summit 2017
60%of mobile connections worldwide are 2G
31. 31
Public + Field: Chrome UX Report
Aggregate
● Form factor
● Effective connection type
● Geography
● UX performance
Analyze
● Raw data on BigQuery
● URL resolution on
PageSpeed Insights
● Mobile Speed Scorecard
● Third party integrations
Anonymize
● Publicly indexable
● Origin rollup
● Unique user threshold
● Normalized to 100%
32. 32
Public + Field: Chrome UX Report
First Paint First Contentful
Paint
DOM Content
Loaded
Onload
this talk is brought to you by the number 80
Let’s start with a statistic
known as the “Golen Rule” of web performance
80% of load time is on the front end
after the HTML page has loaded
JS load/parse/execute, image loading, CSS loading/painting, ads loading
if a site loads in 5 seconds, only 1 second is spent on the backend
coined by Steve Souders in 2012
“godfather” of the field of web performance optimization
BEFORE client side rendering became popular
why does it matter that a page loads in 5 seconds?
what’s the different if it loads in 5 vs 10 seconds?
kinsie’s talk: analog vs digital user experience of consuming news content
imagine if a newspaper “loaded” similar to a web page:
flash of unstyle text, image is last to load, ads pop in and reflow the layout, etc
why do we tolerate that kind of experience on the web?
Performance is correlated with business metrics (revenue, conversions)
Many case studies from notable companies
Causal relationship or at least correlation between performance and business metrics
There is a real dollar value to performance
Using trends from Google Analytics
Apply a function to estimate difference in revenue
Given some hypothetical speedup
Every site is different, just an estimate
Generally, faster = better conversions
RUN YOUR OWN A/B TESTS!
PUBLISH YOUR FINDINGS! (wpostats.com)
Fast performing mobile sites will rank higher, all else being equal
News carousels won’t be limited to AMP pages for long
will include objectively fast sites
our users are at a disadvantage
their connection to the web is slow
it doesn’t help that web pages are heavy and only getting heavier
how long to load 1560 KB on 2G?
1560 KB / 70 Kbps = 3 minutes
check my math
mobile pages aren’t much lighter than desktop
growing much faster than desktop
if anything, mobile is where page weight matters most
paid data plans
emerging markets
users hate slow performance
no chance for them to sign up for your service, see your ads, or read your content if they’ve already left your site
there is a very long tail of slow performance that users experience
35% of desktop onloads and
52% of mobile onloads are
slower than 3 seconds
performance matters
so how do we measure it?
each tool answers the following questions differently:
what is being measured?
how big is the dataset?
great for understanding how the page was built (one page at a time)
not so great for understanding how users actually experience it (by design)
the “inexcusible” performance must-haves
cold page load (no cache)
a good upper limit for bytes and requests
fast network
time not very reliable
highly dependent on user’s demographics and app state
what was requested
when was it requested
how long did it take: dns, tcp, ssl, server response, network response
in what order? did it block other requests?
great for understanding how users experience the site as whole
RUM doesn’t really care about how the pages are built - it just measures the users’ experience
great for understanding how users experience the site as whole
RUM doesn’t really care about how the pages are built - it just measures the users’ experience
in order to understand the ecosystem, we need tools that scale up to hundreds of thousands or millions of samples
the chart we saw earlier of the median page weight is from HTTP Archive
public website of high-level stats and trends about the web
people use it for all kinds of things
sobering observations about the state of the web
academic research
standardization support
the data of the median page load time is from the Chrome UX Report
collection of real user performance data around the world
first we have to identify which sites are built with WordPress
we’ve recently started using Wappalyer to tell us not just about WordPress, but over 1000 other web technologies
43% of 462K
higher than the anecdotal “30% of the web” stat
biased towards the head 500K?
possibly also some false positives (implied detections in Wappalyzer)
compare WordPress sites’ stats against the global stats
page weight is the total number of bytes loaded on the page
median page weight for WordPress is 1921.8 KB
it’s 190.8 KB bigger than the median web page
WordPress sites load a lot more stuff
we can see the how this stat has changed since we were first able to detect WordPress in 2016
WordPress sites have grown by 19.1%
the growth is slightly slower compared to everyone
we can drill down to the page weight for specific resource types, like CSS
unnecessary bytes sent over the network
at the median, it’s better than the web as a whole
~27% of 2.9M
https://bigquery.cloud.google.com/savedquery/920398604589:3eb540c201c442139b42301a845f19e5
The Event Studio
Alley Interactive
Dekode
Inpsyde
Reaktiv Studios
Trew Knowledge Inc.
Ndevr, Inc
YIKES, Inc.
Rinat Khaziev
Companies manage many more websites than just those for their corporate home pages
Compared to other known WP sites, the density of fast FCP is about the same
But attendee’s VIP sites have much fewer “slow” FCP experiences
A few very high performers
Most VIPs are in the “average” range (1-3 seconds)
There are three trends that become visible (the longer you stare at it)
the pages with the smallest density of fast FCP are the heaviest
the pages with the biggest density of fast FCP are the lightest
everything in between generally has a lot of stuff
this makes sense
median wordpress page loads more stuff than the median page
average wordpress performance is worse than average performance
we can’t improve what we don’t measure
i will dedicate my time to ensure that we have the resources to measure performance and our efforts to improve it
individual site owners like you need to monitor your own performance
incorporate it into your culture
make a habit of asking “ok but how fast is it?”
celebrate success studies
publish case studies of perf improving conversions, etc