Comment by nindalf | Hacker News Reader

Comment by nindalf | Hacker News Reader

Parent Full thread

The article talks about Core Web Vitals in passing. That's the major change here. Two posts from May 2020 talk about them more

- https://blog.chromium.org/2020/05/introducing-web-vitals-ess... introduces these metrics, what they mean and how to measure them.

- https://developers.google.com/search/blog/2020/05/evaluating... talks about how the search engine experience will change

> The change for non-AMP content to become eligible to appear in the mobile Top Stories feature in Search will also roll out in May 2021. Any page that meets the Google News content policies will be eligible and we will prioritize pages with great page experience, whether implemented using AMP or any other web technology, as we rank the results.

> In addition to the timing updates described above, we plan to test a visual indicator that highlights pages in search results that have great page experience.

Seems like a positive change. It will mean extra work as developers improve the performance of their sites. But having clear metrics to improve will make that work tractable. Also, advocating for that work to senior management will be easier when it's so clearly tied to SEO.

The upshot is that ordinary users will experience a more performant web. Not overnight but over a few years, like the shift to HTTPS and supporting mobile web versions. Both of those changes were driven in part by the desire for better ranking on Google.

Yeah, but in reality most news sites will fail Core Web Vitals in their current state.

Out of 71 tracked articles on news sites, only 3 or 4 score 85 or higher in overall Performance as tested by Google Lighthouse.

Article Performance Leaderboard: https://webperf.xyz/

good

the amp detractors have always said that we don't need amp because authors can just make their websites faster instead. here's the chance to see whether or not that's true.

news organizations have generally terrible performance and deserve to be punished for it in search rankings.

It is very true! Giants might move slower (as always) but most ad-powered websites have been working like headless chicken to get these metrics in the green.

Granted, it's hard so some may be satisfied with orange metrics.

But I've seen on all publisher-friendly communities I am in how much of an Earthquake the CWV Google Update has been, even if it is now pushed back.

Let's hope more and more follow that trend because nobody hate a fast-loading site with good content!

> even if it is now pushed back

It was pushed back by 6 months, but it's live now.

I meant to say Google using CWV as a ranking signal being pushed back, apologies.

With AMP they just offload their concerns to Google, and I assume they will stay doing that. They have no alternative to their ridiculous ad-tech burderns so I can't see it changing.

But who this does help is websites who are capable of producing a fast experience and don't rely on a dozen ad-tech companies snippets, which will tend to be the smaller sites.

There is still an incentive for publishers. AMP typically showed lower CPMs on ads, and it is possible to have faster loading pages than AMP with ads. Just look at DotDash and their range of verticals.

I personally hated amp, because then I got version of page without dark mode and with limited content instead of real one.

I am actually fine waiting 200ms longer to get those.

It was often 10000ms for me, and there would be popups and paywalls that AMP circumvents.

It's fascinating to me as I moved away from Chrome on mobile specifically because amp caused me so many problems. The solution to page load failures was always editing the URL to give me the original which worked so much better.

Hear that Google? The experience was so bad that I changed browsers to avoid it.

AMP actually supports paywalls: https://amp.dev/documentation/components/amp-access/

This extension is also actually kinda interesting in that it embeds a small domain specific language [1] that's interpreted with a lex-style parser in JavaScript.

This can itself be a bit heavy on the CPU if a website uses it too judiciously!

[1] https://github.com/ampproject/amphtml/blob/main/extensions/a...

> news organizations have generally terrible performance and deserve to be punished for it in search rankings.

Most news on the internet is plagiarized off of a few original sources at best and then redistributed on various sites and portals.

This is a devastating problem for a democratic society. For example, local reporting in the United States has been eroded to the extreme. When a society does not have a commitment to factuality, people become more susceptible to fake news (propaganda), conspiracy theories (such as QAnon), and tend to become more uneducated. Yes, it is true, education can be very easily undone, including in highly intelligent, well-educated individuals.

Facts do not come naturally, nor do they come out of “progress”, such as technological progress. Factuality requires hard work and a lot of money, and it requires reporters on the ground, including in foreign places, such as in the case of foreign correspondents. Unfortunately, the use of foreign correspondents for American news services, physically on the ground in foreign lands, has almost been completely eliminated.

The lack of foreign correspondents means that when a crisis abroad occurs, Americans often have to fill in the gaps and infer what is going on. This allows bad actors to take control.

Of course people do not like to hear the facts. They are hard to listen to and believe in, but they are healthy, especially for a democratic and free society. Of course they are healthy for you as an individual.

The Financial Times has excellent reporting with foreign correspondents located in various countries across the world. The Financial Times also has great tech reporting. In the 8 years that I have been a subscriber, I have only found myself disappointed while reading one single article reporting on tech. Also, of course, I subscribe to local newspapers too.

>news organizations have generally terrible performance and deserve to be punished for it in search rankings.

I wonder how this affects that Australian law about having to tell news sites about algorithm changes, etc. or maybe is affected by.

Is that law in? Being a Googler I kind of can't imagine search working well under it. Would be an absolute spamfest.

It is in place I believe, Google made some concessions and workarounds.

https://www.legislation.gov.au/Details/C2021A00021

https://about.google/google-in-australia/an-open-letter/

Everyone loses, frankly. It was one monopoly versus another, and the Australian people were never a concern.

That reads as "we agreed to send some money", not "we agreed to delay ranking changes by 4 weeks". The latter is the interesting part, as it is incompatible with the reality of needing thousands of algorithm updates per year.

Edit: ah, there's now a provision for this:

> (ii) if the change relates to a matter of urgent public interest—no later than 48 hours after the change is made; and

As much as I hate AMP (well, google specifically), you're right. News orgs will never, ever fix their heavy ass webpages unless they're forced to.

> news organizations have generally terrible performance and deserve to be punished for it in search rankings.

Shouldn't the quality of a site's _content_ count more than its performance?

All else equal, it should. Delivery matters too though, since inaccessible content isn't really useful content

Obviously Google search isn't going to give you an article about Oranges when you query for Apples, but when there's a ton of articles about Apples, having a website with better performance will rank higher. It's unclear how the knobs are tuned, Google's ranking algorithm definitely has a lot to it.

All that is a bit irrelevant, this whole thing is about the news carousel specifically, which is basically showing a bunch of articles in a carousel at the top of search results. Previously only AMP articles showed, now I'm guessing you need some minimum threshold of performance. Seems reasonable, you don't want really slow pages that would push away users.

A simpler explanation.

Google will eventually ban any big enough ad vendor tinkering too much with delayed loading.

They, obviously, cannot ban themselves.

>Google will eventually ban any big enough ad vendor tinkering too much with delayed loading.

Delayed/deferred loading is supposed to improve page performance metrics, not degrade them.

In light of this, I fail to see how you arrive at this conclusion after an article that essentially says that Google decided to de-prioritize AMP in search results and instead give top spots to well-performing websites regardless of whether they are AMP or not.

If anything, this move encourages delayed/deferred loading for all non-AMP websites, because that's one way to improve your website performance and get your search ranking higher.

I my experience google habitually bans just anything seemingly tinkering with ad loading code

Most of the junk at the bottom of the list combines hostile web development practices with criminal negligence of good journalism. If nobody ever visits sfgate.com again, that will be a benefit to humanity.

But despite its name, AMP isn't much help to improve your website's speed though, for instance the new Reddit website uses APM and isn't faster[1] than the old one by lighthouse's metric (and I find it significantly slower from a user's perspective).

Semi-unrelated trivia: Google lighthouse's own website is a disaster[3] by their own standards (with a score of 28), which I find pretty ironic.

[1]: https://developers.google.com/speed/pagespeed/insights/?url=...

[2]: https://developers.google.com/speed/pagespeed/insights/?url=...

[3]: https://developers.google.com/speed/pagespeed/insights/?url=...

> for instance the new Reddit website uses APM and isn't faster than the old one by lighthouse's metric

The AMP page will still load faster because it conforms to a spec which is known to be preload-safe. This means it can be served by services like search engines without any additional network activity, and with minimal layout calculations needed.

Ultimately that's what AMP was designed for. It's more than just a head-to-head speed comparison.

As a sidenote though, reddit's AMP implementation is horrendous for a dozen reasons. It's almost impossible to escape loading the real site, which is not at all within AMP's design guidelines.

> preload-safe.

> served by services like search engines without any additional network activity

You mean AMP pages egt preferntial treatment by Google, and all AMP-related Javascript (IIRC, almost 1 MB of it) is loaded the moment you search anything through Google.

When you hit an AMP page, that JS is already preloaded and, true, there's "no additional network activity".

I'd love for Google to serve my pages' Javascript as well when I search something, and get preferencial treatment, but alas.

> You mean AMP pages egt preferntial treatment by Google

Promoting pages in a carousel above-the-fold is preferential treatment. Preloading Amp pages however is not. This capability works with any implementation of an Amp Cache, including the one used by Microsoft's Bing.

In which world is preloading javascript needed to run a proprietary superset of HTML when you search for something not preferential treatment?

Ah, I guess preferential treatment by Bing makes it alright. If we forget for a moment that Google has 92% market share among search engines.

Everyone always brings up Reddits shitty implementation whenever they want to complain about AMP. AMP was never made for the kind of content reddit serves, it's more for static content like news articles. Reddit's mobile web version is shit in all sorts of way, and it's clear that's intentional to push people towards the app. You can tell by the huge damn banners that show up every single time blocking the content, no matter how many times you dismiss them.

reddit lets you choose freely between various frontends, I only ever use old.reddit.com, there is also i.reddit.com which as far as I understand is an alternative mobile frontend.

But unfortunately almost all Google links, which is how I usually end up on the web version, lead to the amp version. On desktop I may have an extension to automatically switch to old, but on mobile it's quite annoying.

Is the suite that you use to manage these tests available on a git repo by chance?

I really like how transparent you've made: https://docs.google.com/spreadsheets/d/1sGKmbnW74u9r1GOzAQcI...

And I wanted to run something similar but for our own network of sites. (If so you can reach me on my email in profile). Have about 400 sites to access.

How is that data compiled? Just poking around, the worst site in the list (SFGate) seems to have gotten a "1" for performance every time it's been tested, but when I try checking the same link in lighthouse (mobile mode) it scores 55~60.

Uses Google Lighthouse. Takes average of last 3 days worth of tests. Each site usually tested 1-2 times per day.

All the data is here:

https://docs.google.com/spreadsheets/d/1sGKmbnW74u9r1GOzAQcI...

That's the data I asked about. That data shows SFGate repeatedly getting a 1, but when I try the same test it gets 55-60?

There is no one lighthouse number. The site is likely using their own scores that are based on (but not direct averages of) lighthouse. Perhaps they give a score of 1 to the lowest site and scale up.

We're talking about the numbers that show up in the results of a lighthouse "performance" test. The overall score, total blocking time, time to interactive, etc. The spreadsheet has columns for all those values, and it shows SFGate consistently getting a "1" as its overall score.

(It's "0.01" in the sheet, as I guess it considers the scores percentages.)

CWV metrics are gathered and aggregated from field data with the variety of devices and internet speeds you would expect in the wild.

I was asking about the lighthouse results in the link in GP's post, is that what you're answering?

Oh, my bad! Last time I checked, Lighthouse in Chromium used (by default) a Moto G4 on a mobile network simulation.

Would it be possible to add tech news sites to the list? I see you've the onion, but no engadget or ElReg.

Or just Google's own products/services. They usually fail. Same with Apple, Microsoft, everyone. I don't like the "Core Web Vital" metric because it is possible to make your site load slower or in a non-pleasing way for users and improve your score.

I don't understand how this isn't a rehash of Google's PageSpeed push during early 2010s. I do remember having to do the same kind of procedure to get better SEO: measure against PageSpeed's metrics set and optimise the bad results.

This looks just the next step after PageSpeed and I have no idea why Google didn't push this before the whole debacle with AMP, such stupidity but expected nowadays from the tech giants...

In the 2010s, PageSpeed Insights' metrics compared to your site to how it could be if it was delivered optimally. Did you compress your images? Have you combined your CSS files? Are your static resources longcached? A huge site perfectly delivered would score high, while a tiny site sloppily delivered would score low. They represented how much room for improvement your site had if you didn't want to make deep changes.

Search was doing its own thing, and they didn't ever have any public speed metrics.

Core Web Vitals is a different approach: pages are compared to thresholds. A huge site perfectly delivered now likely scores lower than a tiny site sloppily delivered. (PageSpeed Insights now reports CWV metrics.) These are the kind of metrics you can use for search ranking, because they are about user experience instead of developer room for improvement (low-hanging fruit).

(I used to work on mod_pagespeed, which would automatically optimize sites, and I still work for Google. Speaking only for myself.)

Thanks for the clarification here, but I wanted to confirm that when you're talking "huge site" and "tiny site" that this referring to the overall page load size and not the number of pages on the site?

Yes, sorry, I should've said "page" instead of "site"

Awesome. Thank you!

They've done a stunt like AMP before too, when they tried to get everyone to use their hashbang standard.

s/stupidity/greed/

You know it's true.

Some of the metrics they judge are things that AMP basically implements for you, and are a pain to implement otherwise.

Cumulative Layout Shift is one of those things. Content blocks on the page need to have a fixed height, not one that is dynamic (which might happen with lazy-loaded content).

For some use cases, conditionally loading content (one of those being ads) becomes difficult/impossible if you're using a third party system and can't render server side.

Reply on news.ycombinator.com