Last week I enjoyed the opportunity to attend O’Reilly’s Velocity Conference in New York with the team from Rigor.
If you’ve never been to Velocity before, I can tell you that a fellow attendee summed it up accurately as, “[Your] brain feels sore, like [your] jaw after the dentist — from keeping it wide wide open so tons of stuff can be shoved in it.”
When the conference wraps up, all of the new information is still settling in. After having the weekend to process, here are the things that stuck in my mind as important takeaways from this year’s Velocity Conference.
AMP isn’t a silver bullet.
Going into the conference I didn’t know much about AMP beyond, “AMP articles load super fast on my phone.” I learned that AMP forces a set of strict rules on site content and uses Google’s high speed image caching CDN to improve performance. AMP has plans to open up additional functionality – including support for forms – and is working on tools to improve caching to avoid serving up stale images.
While you can easily use proactive synthetic monitoring to monitor the performance of an AMP page, you can’t install traditional RUM tags to collect real user data on an AMP page. In fact, the ability to collect RUM data is built-in with AMP and you can use AMP’s built-in beacons and analytics tools to pull data out into dashboards for reporting. Some RUM providers will leverage these tools to pull AMP data into their products or existing dashboards. For sample pages using real user data, researchers found that AMP could make a page 3.8 faster or 5.7 faster on mobile.
Implementing AMP can be a wonderful way to streamline content, remove third-party bloat, and deliver a faster, better user experience. It’s especially effective for media articles or product pages. But the bottom line is that you could achieve similar or better results by implementing your own optimization and web performance best practices. Even leaders from the team at Google acknowledge that AMP wants to bring the web up to speed, but doesn’t expect to be the “fastest car in the race” and “if you can build an entire web app with AMP, [they] have failed.”
See the slides from AMP: Does it really make your site faster? presented by Nigel Heron and Nic Jansma.
Sites with less script tags have higher conversion rates.
One of the things that makes AMP so effective is that it reduces third-party bloat and the number of requests required to load a page. Reducing the number of script tags on a page doesn’t just improve performance; it also improves conversion rates.
See the numbers in the slides from Using machine learning to determine drivers of bounce and conversion presented by Tammy Everts and Patrick Meenan.
People aren’t more tolerant of slowness on desktop compared to mobile.
At Velocity Conference I was excited to have one of my favorite web performance myths debunked. I’ve always heard and repeated like a parrot that “people were less tolerant of slowness on mobile.” Turns out, the data shows that people behave similarly on desktop and mobile, meaning we’re impatient for slowness on all device types.
This was another great point brought up by Tammy Everts in this presentation Using machine learning to determine drivers of bounce and conversion with Patrick Meenan.
Web performance requirements should influence design decisions.
Often our workflow is to design an ideal, immersive, interactive web experience and then try to figure out later how to make that experience happen within the constraints of our web performance requirements, applying fixes and optimizations after the fact.
As Mark Zeman with Speedcurve illustrated, those same web performance constraints can be powerful drivers of creativity when considered and shared at the onset of a project’s design. In one example project, the design concept called for a page of tiles of streaming video that would expand on interaction. Considering requirements for running multiple video streams at once, the team made a design decision to film the video sequences in stark black and white with high contrast, which not only made the streaming video files smaller but led to striking interactive art in the browser.
As auteurs of the web, we should consider performance requirements during the design process because those constraints can lead to beautiful results.
Watch an earlier presentation of Mark Zeman’s talk from Velocity Amsterdam 2015.
Log code changes that might affect performance.
In his talk, Jim Pierson from GoDaddy mentioned an example scenario where managers reviewed performance trends in a weekly meeting and if they spotted a problem it was “too late” to go back and figure out exactly what caused that change.
Don’t let this happen to you! Log your changes in your proactive monitoring system so that you can compare performance impact before and after events.
See the abstract for Jim Pierson’s presentation here.
If you’re using an internal chat tool like Slack of HipChat in your organization, take advantage of bots to help streamline operations and improve visibility and transparency around who’s owning a particular issue. They can also be especially helpful for teams to audit changes, because everything can be logged in the chat history.
At Rigor we’re already using bots and webhooks to log performance tickets, receive and respond to operational alerts, and track changes that make our sites slower or faster after deploys. If you haven’t implemented some type of ChatOps into your workflow, now is the time; the future is here.
Watch Ashish Kuthiala’s Wall-E-esque keynote here.
Embrace other humans.
I was at Velocity Conference for less than an hour before I came down with a serious case of the warm fuzzies.
My experience at Velocity was friendly, positive, and inclusive. Speakers were constantly tying technical problems and systems back to real humans. Whether it’s making sure you don’t wake someone up in the middle of the night, making sure someone can safely get a ride from their phone, being the CEO who thanks your team for configuring the servers correctly this year, or ensuring that mobile visitors on 2G networks can get critical information in the world’s most remote locations, it’s clear that people working in web performance and DevOps care about other people – even if some of us are introverts who feel more comfortable connecting online than in real life.
But embracing other humans isn’t just about feeling the warm fuzzies. Data shows that gender-diverse teams who amplify ethnically-diverse voices return the best results. We can apply DevOps methods to create better collaboration and be more effective within our companies. And the technology, models, and systems that we develop will only become stronger as we continue to welcome more different voices and bridge gaps between groups.
What were your takeaways?
Velocity Conference breaks talks intelligently into tracts, so it’s impossible to see everything. I’m still catching up on talks, reviewing notes, and reading other post-Velocity blog posts for more ideas.
So, what did I miss? What were your favorite talks from Velocity New York? Let us know on Twitter @TeamRigor with the hashtag #velocityconf.