Let’s examine the difference to your users (and your server bill) between various image loading strategies. We’ll compare “standard” synchronous loading vs. asynchronous vs. lazy loading with the new IntersectionObserver
API.
The modern web is heavily dependent on images. I don’t think that’s news to anyone. As such it’s worth considering your strategy when loading heavy content such as images or video.
Sending content that is too heavy (in terms of MB) is bad for many reasons, here are 4:
- Slow down your website performance.
- Annoy your users – your site is slow and it’s eating all their bandwidth.
- Waste your money – you’re spending money on unnecessary infrastructure / third parties just to deliver and support your massive files.
- Destroy your search engine rankings.
As with most things, there’s a balance to be struck. Usually this is a tripartite discussion between marketing (who want the highest possible content quality), web developers (who want the lowest allowable quality to improve page load speed) and business (who want to make money & don’t necessarily care how that’s done).
How do we strike an acceptable compromise between these three competing interests and ensure that you’re getting the best balance? Here are a few steps which I’d attack in this order:
Resize Your Images
Modern digital cameras will produce images thousands of pixels square. 99% of the time these are inappropriate for directly uploading to a website. Your screen is a thousand pixels square at most. If you’re know the physical pixel size of an image (say 250x250px for an icon) – why bother loading a 2500x2500px image?
You’re sending (and paying for) 10x the amount of data to be sent to the user. The user’s browser then has to spend time resizing it down to 250×250 anyway. A total waste of time.
Reduce the physical size of your images wherever appropriate. Especially for mobile-focused websites.
Optimise Your Images
There are various online websites offering image optimisation these days. One I particularly like is TinyPNG.
Using this image from Adrien Ledoux, TinyPNG shrunk the image from 1.2MB
to 367KB
(a 70% reduction) and I can’t see any difference in quality.
Sites like TinyPNG can shrink images by as much as 90% with no discernible loss of quality.
Load Your Images Appropriately
So, by now your images are an appropriate physical (pixel) size and you’ve optimised them to get the best balance of quality and size.
How you deliver images to your end users is critical. Let’s investigate 3 possibilities…
How you deliver images to your end users is a critical factor. It’ll save you money, improve your search engine rankings and make for happier customers.
Synchronous Loading
This is the standard, default solution. Just include an <img>
tag on your page. The page will load top-down, in order, which means large images will block the rest of the content (text) until they’re finished.
I created a large (13.6MB) dummy page, loaded it & recorded the timings from Chrome DevTools:
Connections | Size | Finish | DOMContentLoaded | Load |
---|---|---|---|---|
7 | 13.6MB | 4.18s | 356ms | 4.01s |
We used almost 14MB of data to load this page and it took over 4s.
Asynchronous Loading
We load a default “loading icon” for each image. Then, in separate threads (without blocking the loading of the page text), we load the images.
This improves loading time of the content (arguably the most important part) and delays your “lower value” assets (images).
However, you still load all of the images, whether or not the user actually ever sees them.
Connections | Size | Finish | DOMContentLoaded | Load |
---|---|---|---|---|
8 | 13.6MB | 4.31s | 267ms | 443ms |
We’ve improved Load time by 89% but still loaded 14MB of data.
Lazy Loading
This solution uses the new IntersectionObserver
API which intelligently “knows” what the user can see (ie. what content is currently within the visible browser window).
Only when the images are actually visible are they loaded. This saves a huge amount of bandwidth for you and your end users.
Connections | Size | Finish | DOMContentLoaded | Load |
---|---|---|---|---|
5 | 3.9MB | 1.52s | 229ms | 401ms |
Huge improvement here for page size. We’ve saved almost 10MB of data.
Conclusion
There is no single “best practice” - each website is unique and will most likely demand a mixture of these techniques.
The key here is that there are a range of techniques to choose from and you owe it to yourself (and your users) to investigate each one and use the most appropriate for you.