Optimizing Images on Alexa's Top 100 Global Sites

Optimizing Images on Alexa's Top 100 Global Sites(blog.kraken.io)

35 points by rast-a 12 years ago | 12 comments

dm2 12 years ago |

A third of Google's image size is this thing: https://ssl.gstatic.com/s2/oz/images/notifications/spinner_3... which is basically a loading icon that is rarely seen.

How does HTTP compression factor into image optimization? Is it possible to optimize an image but the HTTP compressed size ends up being greater?

You should release a monthly report with this same algorithm. You should also put in a column for competitors algorithm (not the just ones you beat, all of them).

Don't forget that some images are loaded after the page and content load, so they have very little impact on user experience (and all these sites make heavy use of CDNs).

thrownaway2424 11 years ago | |

That obnoxious spinner is used for Mr. Tinkles, the Asshole Of Notification. Because even after Mr. Tinkles starts ringing out, he has no idea what to say. So when you click on him to make him shut the hell up, he still hasn't loaded the hundred bytes of your notification, and you have to look at that stupid spinner while all of Google's horrifying social storage infrastructure lurches into action. And because Mr. Tinkles is part of Google's all-in Social Death Pact, you have to load that image on every Google property: Gmail, Maps, Drive, Search, everything.

yalooze 12 years ago | |

At first I assumed Google's reduction was so high because they'd scraped the site on a day when the logo was a doodle. But I just manually checked Google's main logo and was surprised to see a 40% lossless reduction. I would assume they dogfood PageSpeed (or at least use some internal version of it) which does image optimisation automatically for them. So is Kraken just better or is it just an oversight from Google?

dm2 12 years ago | | |

There might be a browser support issue that prevents doing it.

The bandwidth and mobile loading might not be an issue because of their massive amount of data-centers to serve content from.

It seems like it to would be easy to test the new image to see if anyone reports any issues, but there might just not be any benefits to reducing the homepage image by 40%. Make sure you consider that the homepage image is being scaled smaller than the actually image is, some optimizers take this into account and will give you a scaled image.

tmikaeld 12 years ago |

Hm, tested the field/mountain comparison from kraken.io first page on JPEGmini - 184kb on jpegmini vs 242kb on Kraken.io

Seems their solution is both cheaper and better?

http://www.jpegmini.com/

martin-adams 12 years ago | |

I just tried two product images on both and JPEGmini was either the same filesize or larger. So I think the subject will have something to do with it.

ksec 12 years ago | |

Well the server version start at $199. So i dont think that is cheap.

tmikaeld 11 years ago | | |

Considering you are not limited to any amount of images, yes it is.

BorisMelnik 12 years ago |

I don't know much about the technical side of things, but as someone studying SEO and UX this seems like a huge win. Not only by cutting down on page load size / speed but so much better fo rUX.

noir_lord 12 years ago | |

> I don't know much about the technical side of things, but as someone studying SEO and UX.

I felt a great disturbance in the Force, as if millions of developers suddenly cried out in terror...

Joking aside it is usually a win to reduce page load sizes from a speed and bandwidth point of view (which on a heavily tracked site can add significantly to costs) occasionally you have to be careful as some of the compression methods result in non-standard or "technically standard but the client doesn't really do it that way" files which can render corrupted or more slowly.