Hacker News Reader

Comment by Veratyr | Hacker News Reader

In this case you're probably right but in general, latency can make a big impact on profit and revenue.

For example Marissa Mayer claimed (though this was in 2006) that a 500ms delay directly caused a 20% drop in traffic and revenue: https://glinden.blogspot.com/2006/11/marissa-mayer-at-web-20...

Optimizely found that for The Telegraph, adding a 4s delay led to a 10% drop in pageviews: https://blog.optimizely.com/2016/07/13/how-does-page-load-ti...

100ms cost Amazon 1% of sales: http://blog.gigaspaces.com/amazon-found-every-100ms-of-laten...

Akamai has a number of claims: https://www.akamai.com/us/en/about/news/press/2009-press/aka...

Of course the impact of latency is going to depend on your particular circumstances but there are certainly circumstances where it can make a big impact.

View on HN

Google: 2006.

Amazon: 2006.

Akamai: 2009.

At least the Optimizely result is from this year.

I really want more evidence of the importance of performance in web systems (to justify my hobby, if for no other reason), but things from ten years ago just don’t cut it.

Can we please stop citing articles from ten years ago and cite some studies that aren’t more than two years old?

Do people have more such studies?

(Also I’d be interested in more geographically diverse results. Figures like “three seconds to abandonment” are absurd when you’re in Australia or India as I am. Most e-commerce sites—especially US/international ones—haven’t even started drawing at that point, even with the best Internet connectivity and computer performance available, simply because of latency if the site is hosted in the US only. I visited the US a couple of years ago and was amazed at just how fast the Internet was, simply due to reduced latency.)

Why would you expect that visitors became more accepting of slower websites?

If anything I would imagine to be the opposite with connections generally being faster these days. Especially since old Google's result was about observed effects of 100ms slowdown so not about generally slow websites.

I’m not trying to say it will go one way or another—just that the variables in the equation have all changed (expectations, the nature of access to the Internet, ubiquity, &c.), so results from several years ago are obsolete (I’d even call results from over a year ago obsolescent) and we need new figures for credibility.

People don’t necessarily become more demanding; maybe in the earlier days they said, “I don’t need this, it’s taking too long, I’ll just give up,” whereas now they are resigned to the Internet being slow.

Also of interest would be making a site massively faster (e.g. by an apposite service worker implementation), but A/B test on the rollout of it. Over several months, ideally.

> I really want more evidence of the importance of performance in web systems (to justify my hobby, if for no other reason), but things from ten years ago just don’t cut it.

> Can we please stop citing articles from ten years ago and cite some studies that aren’t more than two years old?

Could you explain why? New data is good, sure but why should we disregard old data? I think instead of asking for people to simply discard old data, you should point out the issues you believe the data has so that they can be addressed (with newer data if necessary).

Anyhow, from https://riteproject.files.wordpress.com/2014/10/fact-sheet.p...:

- "In 2008 the Aberdeen Group found that a 1-second delay in page load time costs 11% fewer page views, a 16% decrease in customer satisfaction, and a 7% decrease in searches that result in new customers."

- In 2009 Microsoft Bing found that introducing 500ms extra delay translated to 1.2% less advertising revenue.

I also found a whitepaper from 2010 done by Gomez.com that shows how speed impacts various metrics: https://github.com/JasonPaulGardner/po4ux/blob/master/InfoPa...

There's also an article from 2013 that's only loosely related but shows that reducing TTFB latency can increase search result rankings: https://moz.com/blog/how-website-speed-actually-impacts-sear...

> Also I’d be interested in more geographically diverse results.

I'd love these too, let me know if you find them.

The situation has changed. The nature of people’s consumption of the Web is different now, especially with the take-up of mobile browsing in the last couple of years. (The change in the nature of consumption on that count in particular is why I would like to see results from the last year, especially if they’re segmented by device.)

People depend on the web more than they did, so they might be willing to wait for things longer—or even just be resigned to the web being slow. Or perhaps the increased number of providers means that they won’t be willing to wait as long. I don’t know.

It does seem to me that the precious few studies that there are from recent times haven’t shown as severe a drop-off as ones from ten years ago did. But then the ranges of the results are so far out of my experience in non-US countries that I don’t feel I can judge it.

What I do know is that I have had people complain of the age of the data I cited within the last couple of years, when citing the traditional examples of Amazon and Google. Their age and the indubitable ecosystem changes since their time make me a little leery of citing them now, because after due thought, I agree with the objections.

As I say, I want performance to be a commercially valuable factor, because it would justify one of my favourite two topics in IT. I just want us to be using solid, dependable studies of the matter so that we can prove our point without reasonable doubt, and justify expenditure on improving performance, rather than aged studies which are open to quibble.

> People depend on the web more than they did, so they might be willing to wait for things longer—or even just be resigned to the web being slow. Or perhaps the increased number of providers means that they won’t be willing to wait as long. I don’t know.

It is hard for me to disregard those big names and their claims, especially since I have no direct experience in working for any of those companies for me to say otherwise.

On the other hand, in my purely anecdotal world of friends, family, and colleagues, I have never seen/heard of such behaviors. I doubt any of them even realize what 100ms is, let alone abandon their purchase on Amazon because of that delay. Most people can't even react in 100ms to brake their cars.

I really have to wonder if we have causation vs correlation going on.

I think citing data before the wide spread uptake on mobile is non-representative of the modern use of the web.

But it's probably very reasonable to believe that people's expectations of a faster response time has only risen since 10 years ago.

If I had to make a guess whether the same studies performed 10 years ago were performed today, whether the drop in page views would increase or decrease with slower response time, Id guess that today you'd have even fewer page views than 10 years ago.

I don’t believe it’s reasonable to assume that, having presented an alternative view that I find plausible in other comments in this thread. I hope that people’s expectations have increased, but I don’t believe it’s a reasonable assumption.

We need data.

There is lots of recent data out there. The general trend is that those findings from the mid 2000s hold even more true today. People are much more sensitive to latency, while the problem has gotten harder due to last mile flowing through congested air. https://www.doubleclickbygoogle.com/articles/mobile-speed-ma...

The entire premise of Google AMP and Facebook's Instant Articles is predicated on performance. Speed matters.

I tested the impact of page speed on ads about 3 years ago. The CTR impact was pretty dramatic - I can't recall the exact numbers but they were in the order of 50% CTR drop with 500ms delay.

That's obviously specific to ads, but there's clearly still some impact from page speed there.

My personal mental model is this:

People have a threshold to to [1] on how long they stay on your site, say 5 minutes. If you're site is faster, they will see more pages. More pages increases the likelihood that they find something to buy/comment/...

[1] would love this tested.

You can fix that when the product is released. It's called optimization, you don't need to do it upfront.

Check if people like your software and you receive enough €€€, if it's market git. Optimize and make a blog post about it

But sometimes that fix can only be done by switching to another tech stack or framework.

That is, a complete rewrite.

When that's not possible, we have to endure the constantly crashing and slow as molasses web sites like the ones used in universities by students to enroll in courses.

Ya I have worked with the optimize at the end guys before. Really, you are gonna refactor a 100k Java codebase to be fast when its done? Want to know how much easier it is to plan for speed from day 1?

As always, going too far either way is a bad idea. I've worked with "optimise everything ahead of time" people before and it can be an astonishing waste of time.

You shouldn't completely ignore performance all the way through and assume you can sprinkle it on top later, but you shouldn't spend months squeezing the last few ms out of a feature that the client will change / remove when they try it anyway.

> Really, you are gonna refactor a 100k Java codebase to be fast when its done?

Maybe? Depends where the problems are. If the problems are huge architectural ones, no that sounds terrible. If the problem can be solved by later improving an inner loop somewhere and getting things solved then sure.

Speed is user-experience.

People used Instagram at first because it was fast.

And they were quickly bought out for a billion.

I never heard someone say "I use instagram because it's fast"

But how many people don't X because it's slow? I don't browse on my phone's because ads make it too slow.

Speed is part of the user experience, but again it's more complex than that.

Nobody would have used snapchat if it didn't take photos but the site loaded in 1ms. Nobody would use it if it had loads more features and took 2 hours.

Speed is a part of it, and must be traded off with all the other parts of the user experience.

Your particular example made me smile.

I'm currently leading a team whose primary product is the front end service for the core SaaS business that does billions of monthly impressions (hundreds of millions of uniques). This product is over 10 years old, written in Spring, and has around 400,000 lines of Java.

A couple years ago, we underwent a massive overhaul to the system and refactored the application to use our realtime steaming platform instead of using MySQL+Solr.

It took some time to get it done, mainly because at the time there were about 8 years of features baked into a monolithic web service.

Ultimately, we were successful in the endeavor. It took about 2.5 years and our response times are dramatically faster (and less expensive in terms of infrastructure cost).

So yeah, you can do it. It's expensive and takes time if you wait too long. And you need full support from the business. And some companies don't survive it. YMMV.

It's helpful to separate performance from the ability to perform. It's generally not the greatest idea to prematurely optimize, but you should absolutely be making decisions that give you the flexibility to optimize in the future. For instance, adding a caching layer before it's needed may be a bad idea, but organizing your data in such a way that you could easily add a caching layer may be a good idea. Making your website in C because it's fast may be a bad idea, but choosing Python because you could easily add C bindings to the slow parts may be a good idea.

Yeah but if it takes 6 months longer, and is 30% harder to refactor, it doesn't matter. You're not selling anything when it's not running.

This is just moving the goalposts. No one is saying you have to do these optimisations upfront, just that you they are worth doing.

I wonder if Amazon has done other/more research since and softened their stance since that gigaspaces article...

... because I'm noticing, esp. over the last 1+ years, that Amazon's page loads are getting slower and slower (Win7, FF) with all of the "crud" they are loading with each/every page. I remember when it was all a lot snappier (not like an HN page load "snappy", but pretty good!)

Yep they sure put a lot of crap on their clunky pages. They could do much better but they don't seem to care.

my observation is Amazon hardly cares about web page performance..

The 'fast' is not increasing revenue by itself I think. Faster pages => more products seen by user => higher likelihood they buy something.

Amazon has shifted since then to pushing more products into your face, which looks like for them works better than the 'faster pages' approach.

This is reasonable. Aside for must-reads, I don't read articles or news stories, funny posts, memes, gifs, etc. if they take way too long to load (3000ms+), or if it's TLDR (unless it's must-read). Most of the time, stuff ends up being TLDR, and it's a quick skim to make sure I'm not missing must-read. On websites like Washington Post, Bloomberg, etc. there's just too much shit going on. Too many ads. Too many popups. Too many "please give me your email!" requests. Shit's everywhere on the internet. I'm not about to go to Rocket Fizz for a Snickers Bar.

> 100ms cost Amazon 1% of sales

I wonder why we don't have free high-speed internet yet, financed by big companies.

Google tries with fiber and balloons, but it's a very hard market to enter. Telcos are terribly regulated and investment costs are high.

Reply on news.ycombinator.com