Web Performance: Forget Everything I Said Last Time
An updated version of my web performance talk from a few years ago. Here we explain what's changed, what's the same, and how to dive deeper into user centric performance metrics.
WHAT WE DISCUSSED LAST TIME ▸ We need Performance Budgets ▸ Use the right type of images ▸ PNGs can be better than JPEGs ▸ Sometimes all you need is an SVG ▸ WebP All The Things (when it helps) ▸ Measure Measure Measure!!!
WHAT WE DISCUSSED LAST TIME ▸ Why performance matters: ▸ Long load times impacted bounce rate and conversion rate ▸ We knew somehow it impacted the bottom line of businesses selling online ▸ Slow sites suck
SPEED INDEX ▸ Helps us understand how fast the user perceived the page to load ▸ Number of seconds when 90% of the above the fold content loaded ▸ Ability to track “in the lab” for the last few years ▸ Ability to track “in the field” somewhat available
FIRST MEANINGFUL PAINT ▸ Answers the “is it useful?” question ▸ Useful is subjective and hard to measure ▸ “Hero Elements” ▸ Loading these important parts first makes the user perceive the page is loading faster
FIRST INTERACTIVE ▸ Time when the page is both visually rendered and capable of reliably responding to user input ▸ Reasons for long FI: ▸ JavaScript needed to make components on the page work hasn’t loaded ▸ There are long tasks blocking the main thread
CONSISTENTLY INTERACTIVE & INPUT LATENCY ▸ Consistently Interactive ▸ Time at when the main thread is able to consistently respond to user input ▸ Input Latency ▸ Should be 100ms or less for a smooth, interactive experience
“IN THE LAB” TESTING: LIGHTHOUSE ▸ Created by Google ▸ Allows you to measure performance and other important features/analytics ▸ Takes measurements from Javascript and other tools (such as ffmpeg for screen captures)
MACHINE LEARNING ▸ Going to really give us actionable insights ▸ Find correlations between metrics that we don’t realize impact each other ▸ Also a buzz word! (But really it’s just stats and programming)