Too early to get 5.8+ data but to get started here’s a look at WebP adoption by WordPress version as of the April dataset (mobile only):
SELECT
version,
COUNTIF(has_webp) AS pages_with_webp,
COUNT(0) AS pages,
COUNTIF(has_webp) / COUNT(0) AS pct_webp
FROM (
SELECT DISTINCT
url,
info AS version
FROM
`httparchive.technologies.2021_04_01_mobile`
WHERE
app = 'WordPress')
JOIN (
SELECT
url,
has_webp
FROM (
SELECT
pageid,
COUNTIF(ext = 'webp') > 0 AS has_webp
FROM
`httparchive.summary_requests.2021_04_01_mobile`
GROUP BY
pageid)
JOIN (
SELECT
pageid,
url
FROM
`httparchive.summary_pages.2021_04_01_mobile`)
USING
(pageid))
USING
(url)
GROUP BY
version
ORDER BY
pages DESC
| version |
pages_with_webp |
pages |
pct_webp |
|
47,480 |
938,168 |
5.06% |
| 5.7 |
8,212 |
343,103 |
2.39% |
| 5.6.2 |
3,946 |
186,681 |
2.11% |
| 5.5.3 |
1,878 |
111,512 |
1.68% |
| 5.7.1 |
2,316 |
99,861 |
2.32% |
| 5.4.4 |
1,052 |
71,304 |
1.48% |
| 4.9.16 |
569 |
58,499 |
0.97% |
| 5.3.6 |
856 |
46,232 |
1.85% |
| 5.6.3 |
795 |
40,345 |
1.97% |
| 5.2.9 |
601 |
39,449 |
1.52% |
| 5.5.4 |
443 |
30,668 |
1.44% |
| 5.6 |
606 |
24,580 |
2.47% |
| 5.4.5 |
370 |
23,259 |
1.59% |
| 5.1.8 |
165 |
20,615 |
0.80% |
| 4.9.17 |
172 |
19,366 |
0.89% |
| 5.4.2 |
271 |
15,531 |
1.74% |
| 5.3.7 |
237 |
15,435 |
1.54% |
| 5.2.10 |
208 |
13,431 |
1.55% |
| 4.8.15 |
93 |
13,349 |
0.70% |
| 4.7.19 |
83 |
13,299 |
0.62% |
| … |
|
|
|
There are a lot of pages without a discernible version and 5.1% of those have a WebP image. The most popular detectible version of WordPress is 5.7 and 2.4% of those 343K pages have a WebP image.
If we remove the version grouping and look at any WordPress page, WebP adoption is at 3.1%.
Here’s a slight modification to look at the median image bytes per page segmented by those with and without any WebP:
SELECT
has_webp,
APPROX_QUANTILES(bytesImg, 1000)[OFFSET(500)] / 1024 / 1024 AS median_img_mbytes
FROM (
SELECT DISTINCT
url,
info AS version
FROM
`httparchive.technologies.2021_04_01_mobile`
WHERE
app = 'WordPress')
JOIN (
SELECT
url,
has_webp,
bytesImg
FROM (
SELECT
pageid,
COUNTIF(ext = 'webp') > 0 AS has_webp
FROM
`httparchive.summary_requests.2021_04_01_mobile`
GROUP BY
pageid)
JOIN (
SELECT
pageid,
url,
bytesImg
FROM
`httparchive.summary_pages.2021_04_01_mobile`)
USING
(pageid))
USING
(url)
GROUP BY
has_webp
| has_webp |
median_img_mbytes |
| true |
0.64 |
| false |
1.10 |
So the median WordPress page with any WebP contains 0.64 MB of total images while the median page without any WebP contains 1.1 MB of total images. Nice to see that difference!
And as a median percent of all bytes:
SELECT
has_webp,
APPROX_QUANTILES(pct_img_bytes, 1000)[OFFSET(500)] AS median_pct_img_bytes
FROM (
SELECT DISTINCT
url,
info AS version
FROM
`httparchive.technologies.2021_04_01_mobile`
WHERE
app = 'WordPress')
JOIN (
SELECT
url,
has_webp,
pct_img_bytes
FROM (
SELECT
pageid,
COUNTIF(ext = 'webp') > 0 AS has_webp
FROM
`httparchive.summary_requests.2021_04_01_mobile`
GROUP BY
pageid)
JOIN (
SELECT
pageid,
url,
SAFE_DIVIDE(bytesImg, bytesTotal) AS pct_img_bytes
FROM
`httparchive.summary_pages.2021_04_01_mobile`)
USING
(pageid))
USING
(url)
GROUP BY
has_webp
| has_webp |
median_pct_img_bytes |
| true |
39.7% |
| false |
55.1% |
Similarly, WordPress sites with WebP tend to have fewer image bytes in relation to all of a page’s bytes.
Here’s a look at the median lab-based LCP for pages that use WebP vs don’t:
Beware: this query is slow and expensive
SELECT
has_webp,
APPROX_QUANTILES(lcp, 1000)[OFFSET(500)] AS median_lcp
FROM (
SELECT DISTINCT
url,
info AS version
FROM
`httparchive.technologies.2021_04_01_mobile`
WHERE
app = 'WordPress')
JOIN (
SELECT
url,
has_webp
FROM (
SELECT
pageid,
COUNTIF(ext = 'webp') > 0 AS has_webp
FROM
`httparchive.summary_requests.2021_04_01_mobile`
GROUP BY
pageid)
JOIN (
SELECT
pageid,
url
FROM
`httparchive.summary_pages.2021_04_01_mobile`)
USING
(pageid))
USING
(url)
JOIN (
SELECT
url,
CAST(JSON_EXTRACT_SCALAR(payload, "$['_chromeUserTiming.LargestContentfulPaint']") AS INT64) / 1000 AS lcp
FROM
`httparchive.pages.2021_04_01_mobile`)
USING
(url)
GROUP BY
has_webp
| has_webp |
median_lcp |
| true |
7.944 |
| false |
9.046 |
So the median WordPress page with WebP has a lab-based LCP of 7.9s while the median WordPress page without any WebP has a lab-based LCP of 9.0s. Emphasis on lab-based LCP because these are mobile tests that are throttled. The exact values don’t matter much but it’s still valid for us to compare which one is bigger.
I know @patmeenan has been working on integrating page-level CrUX API data into the results but I don’t think that’s available in this crawl yet. But soon we should also be able to look at the real-user LCP experiences and segment by WebP adoption.
I’d love to add something to HTTP Archive: State of Images to track WebP adoption. We can segment by the WordPress lens, although that doesn’t give us version-level granularity.
