Is Cloudflare APO Effective on AMP Pages ?

GTMetrix and WebPageTest results using Cloudflare APO and AMP. Tested from 7 different server locations. TTFB and FCP values. Core Web Vitals.

In an earlier case-study we have seen the benefits of Cloudflare APO on reducing TTFB and FCP times. The Core Web Vitals were also green, when there were no video or display ads. The initial page request is served more quickly to the client, when using the APO feature.

For a free Cloudflare plan, it costs only $5/month. If you are on a slow host, this improves the page load time and optimizes the font resources too. Cloudlare APO is not a magic wand to improve the speed of the site.

Also Read – Website Speed Test with and without PageSpeed Ninja and Cloudflare APO

You still need to have a fast server, cache other resources like CSS, JSS, usage of minification, page optimization plugins are also important. The quicker the page response, visitors are more likely to stay on the web site. This will improve the bounce rate and the amount of time spent on the page.

Google has introduced AMP pages for fast mobile responsive pages. Accelerated Mobile Pages (AMP) is the new technology that improves the loading of mobile pages.

Also Read – PageSpeed Ninja vs SG Optimizer – Which is the Best WordPress Cache Plugin

In addition they have the extra benefit of ranking high in Google search results. Though these are particularly useful for news sites and results appear in mobile carousel results at the top, AMP pages rank well in search engine results.

In order to see the benefits of combination of both these technologies, we have done a case-study to see the effect of Cloudflare APO feature on AMP pages.

Introduction

The test site had around 20 plugins. To serve AMP, a plugin was used. PageSpeed Ninja plugin was used for minification, gzip compression, optimize images and page optimization. Cloudflare free plan was also enabled. APO feature was being used for the past 2 months.

Two important metrics of TTFB and FCP were used in the study. Since Cloudflare APO is effective in reducing these two parameters, their effect on AMP pages was studied for the same. Other Core Web Vitals were also reported.

To test these performance metrics, two speed testing tools were used. GTMetrix and WebPageTest are load speed test results which give a waterfall view of the page load. They also report the page metrics like LCP, FCP, TTFB, CLS, TBT and many others.

Since GTMetrix uses the Lighthouse tool, their results are more close to Google’s PageSpeed Insights tool. Since Cloudflare APO benefits can be seen when the server location is farther from the origin, the same page was tested from different server locations.

WebPageTest reports the TTFB times in their report. They also have a waterfall view where we can see the time taken for the loaded resources. AMP pages are served from Google CDN cache. This also reduces the load time of the page.

Test Results – Without Cloudflare APO

The following page was tested from different server locations around the world.

https://www.wpreviewtips.com/amp-for-wp-pro-review/amp/

GTMetrix Test Report

TTFB and FCP Values

S.NoServer locationFCPTTFB
1San Antonio, TX, USA2.9s2.83s
2Hong Kong, China3.6s3.4s
3London, UK4.3s4.17s
4Mumbai, India3.4s3.35s
5Sydney, Australia8.3s8.16s
6Sao Paulo, Brazil5.4s5.4s
7Vancouver, Canada3.0s2.77s

Core Web Vitals Test

S.NoServer locationLCPTBTCLSGTMetrix Grade
1San Antonio, TX, USA3.6s51ms0.43D
2Hong Kong, China5.5s196ms0.43F
3London, UK5.3s77ms0.43E
4Mumbai, India4.2s229ms0.43E
5Sydney, Australia10.1s363ms0.43F
6Sao Paulo, Brazil7.1s97ms0.43E
7Vancouver, Canada3.4s90ms0.43D

Analysis

Without using the Cloudflare APO feature, the TTFB times were very high. Though AMP pages were being used, the GTMetrix test result shows that TTFB is high as 8.16s in some cases. This is very high compared to the TTFB values using Cloudflare APO. Though there were not much of JS and CSS used on the site, both these core metrics were high between 2.5s to 5.5s.

This could be attributed mainly due to the AMP for WP plugin being used for displaying the AMP pages. The FCP values were also high in seconds. As can be seen in the below table, the TTFB and FCP values were low in milli-seconds when using the Cloudflare APO.

Test Results – Cloudflare APO Enabled

GTMetrix Test Report

TTFB and FCP Values

S.NoServer locationFCPTTFB
1San Antonio, TX, USA0.5s131.2ms
2Hong Kong, China415ms207.3ms
3London, UK354ms85.8ms
4Mumbai, India0.7s236.6ms
5Sydney, Australia1.0s113.7ms
6Sao Paulo, Brazil248ms99.6ms
7Vancouver, Canada277ms96.1ms

Core Web Vitals Test

S.NoServer locationLCPTBTCLSSizeGTMetrix Grade
1San Antonio, TX, USA1.1s113ms0.4321.7KBA
2Hong Kong, China0.8s75ms0.4321.8KBA
3London, UK1.7s76ms0.4321.6KBA
4Mumbai, India1.6s431ms0.4321.7KBC
5Sydney, Australia1.5s210ms0.4321.7KBB
6Sao Paulo, Brazil0.9s66ms0.4321.6KBA
7Vancouver, Canada0.6s102ms0.4321.7KBA

Analysis

The TTFB values reported well within good range. As per Google’s standards, 200ms of TTFB is a good score. Normal sites report a value of 300 to 600ms for the same. Site having more than 1s of TTFB is not considered good.

In the above test from different server locations, we can see that all the TTFB values were between 100 to 250ms. This is very fast compared to not using the Cloudflare APO. This result could be attributed because of AMP pages, page optimization and also using the automatic platform optimization.

The lowest TTFB was from London and the highest from India. FCP values were also good between 200 to 500ms, except for the Sydney server. The Core Web Vitals were also showing good scores for LCP, TBT. But the CLS were somewhat higher. A good CLS score is less than 0.25 according to Search Console metrics. But using AMP pages shows 0.43 which is a little higher.

This can be because of movable elements like font files. Though there were no ads, CLS value was high.

WebPageTest Report

TTFB and FCP Values

S.NoServer LocationFirst ByteFCP
1Dulles, VA0.280s1.065s
2SaoPaulo, Brazil1.424s2.281s
3London, UK0.848S1.567s
4South Africa1.509s2.498s
5Dubai, UAE1.353s3.748s
6Mumbai, India0.497s1.399s
7Sydney, Aus1.296s2.098s

Core Web Vitals Test

S.NoServer LocationSpeed IndexLCPCLSFully Loaded TimeRequests
1Dulles, VA1.389s1.765s0.4192.836s19
2SaoPaulo, Brazil2.729s2.848s0.4196.193s19
3London, UK2.177s2.644s0.4213.891s20
4South Africa3.051s3.414s0.4197.544s21
5Dubai, UAE4.512s4.981s0.41915.408s22
6Mumbai, India1.583s1.665s0.4194.070s21
7Sydney, Aus2.743s3.198s0.4196.135s21

Analysis

As above, the AMP page was tested from 7 different server locations around that world. This test clearly reports the TTFB times, unlike the GTMetrix test. Though there is some ambiguity in the before test, WebPageTest reports the initial page load time as it is.

The TTFB values reported for the AMP pages were between 0.5s to 1.5s. Except the USA server which was close to the origin, all the other values were higher. As mentioned earlier, this test uses a different set of test tools to report the TTFB and Web Vitals score.

So the benefit of Cloudflare APO on AMP pages could not be seen in this test.

The FCP values were also between 1.5s to 3.5s. This is also little high compared to the earlier test of GTMetrix. Since AMP technology is supposed to reduce these values, there was not much benefit as seen in the test results.

CLS values were also higher in the range of 0.419. The fully loaded time also was varying from different server locations. London and Dulles servers were showing less time to fully load the page. Dubai, South Africa, Brazil locations were showing higher page load times.

Speed Index values were also around the 2 to 3s range. These values were higher compared to the GTMetrix test.

4.7/5 - (3 votes)

Leave a Comment