Dave Archer

@TimVereecke If argue visitors with a "good" lcp don't care care about that either. Unfortunately it's a very broadly researched number and won't make sense for all sites

Visitors with a bad LCP experience couldn't care less about the fact that your 75th percentile is actually "good" #webperf #perfnow

Thinking about building something a bit special in terms of user experience monitoring. Take a look here: mailchi.mp/c4ff257ba04a/b… If you like what you see, please do add your email to help give an idea of its usefulness. Also, sharing is caring, so please share!

@obeattie Unit tests as a regression guard fail when we refactor at a higher "unit". Subtleties and edge cases in the now-deleted code and corresponding tests are lost immediately. That's a long way of saying I agree

If I could only have one kind of test, I would 💯% just have end-to-end tests that can run continuously against any environment, checking that the things actual customers want to do work. The feedback loop is longer than unit tests, but the ROI very high.

@hankgreen What is this? A galaxy for ants??!

Everything in this image that doesn't have spikes coming off of it is a galaxy. Every. Single. Dot.

@Steve8708 I don't think it's precisely codebase size that matters, but to be pedantic it's more functionality being added. Poor organisation of execution, network calls, redraws etc. The browsers and networks do what they do pretty well. We do the silly things

A big problem with how we build web frontends today: Your brand new project is fast (woo!), but as the codebase grows, perf declines rapidly We treat this as if it's the developer's problem, but when it's the case for virtually every site, it's clearly a tooling problem

@Steve8708 @____lighthouse I'm biased here, but I'd argue that a RUM solution is better for tracking. Don't get me wrong, lighthouse is amazing, but it runs under controlled conditions. Good for a snapshot check and build pipelines (which RUM can't do obvs)

then, be sure to track your performance always. love it or hate it, @____lighthouse is the gold standard here, and esp. look at the google pagespeed insights *mobile* score for emulated real world metrics (your m1 laptop on 1gbps wifi is not real world!)

Innovation in web performance has been 🔥 lately. Here are the projects you need to watch in 2022: 🧵

@TimVereecke @__treo Point 2 is almost good enough on its own. As for "proof", I've been building a thing to try to do just that bytesmatter.io/blog/performan… (actually wanted to dm first, not trying to advertise, but I do think it is genuinely important)

My changes in the beginning of January will likely result 97+% good LCP… later than expected, but it gets harder and harder to move the needle. Screenshot from @__treo #webperf

Controversial? Purely in terms of #webperf (i.e cost per MB aside), page WEIGHT doesn't matter, but page WAIT does. Weight optimisations aren't always (very) effective because of browser optimisations. Subtle shift to focus on WAIT means a focus on user

@alecmce @EmmaBostian Aw cmon. You just fell back on good old “it depends”.

My personal opinion: Having 100% test coverage doesn't make your code more effective. Tests should add value but at some point being overly redundant and testing absolutely every line of code is ineffective. Ex) Testing that a component renders doesn't IN MY OPINION add value.

@alecmce @EmmaBostian And I just got the dance off reference 🤣

@RaoDaoZao @sarahjane_hale And when will it be ready

@sarahjane_hale When It's Ready.

On the plus side: you can no longer walk off the edge of the map, so this level is now technically shippable. 😌

@TheRealNooshu Gonna use this opportunity to plug my offering that would have given you the exact reason bytesmatter.io/blog/waterfall…

On the 'report COVID test results' page's RUM data, I see 1 session where it took 29 seconds to load the page for a user on a very slow connection. Only way I could reproduce was applying 10% packet loss to the synthetic test connection (that's a lot!). Thankfully this is rare!

@harrisongilmore @BytesPerf your social reach just doubled the vote count

@BytesPerf @furybyname Have a vote! ⬆️

Product Hunt launch day yay! BytesMatter: Web monitoring and analytics through the eyes of your users producthunt.com/posts/bytesmat…

@TheRealNooshu Wouldn’t it be the simplest explanation: batch job running?

RUM data from GOV.UK shows an increase in both backend and frontend performance time at 7am every morning. Now considering almost 92% of traffic is from the UK, is that just general internet congestion as people are all waking up? Or something else? Thoughts?

@harrisongilmore @ilanamunckton Always Dave

@ilanamunckton “This is the third time someone has written this exact same document Dave, just stop it.”

My morning revelation is that tech companies should hire... Librarians 📚 How much time is spent trawling through Confluence, Google Docs, Notion and other doc management platforms? Imagine that was curated by someone who knew what they were doing? IMAGINE

Hope this can help others. We ran into an issue where Nock wasn't recording all requests as expected. "You'll never guess what solution came next"! bytesmatter.io/blog/why-did-n… #SoftwareTesting

"My website is so fast I had spare time to make this!" youtu.be/H2B6F9SFut8 🤣🤣