Investigating Verizon’s Network Optimization

Earlier this year, Verizon Wireless adopted a process it calls Network Optimization with the goal of more efficiently transmitting data to its customers, and thereby reducing the load placed upon its wireless data network. It turns out that this process actually works by intercepting data sent to your phone, particularly HTTP data over port 80, and applying lossy compression before passing it along to you.

Verizon claims that the information it removes from images and videos is conservative, for example by removing only colors not visible to the human eye.

Rather than simply trust Verizon, I decided to empirically test the effects of its optimization process.

The experiment

For my experiment, I built a simple app for my iPhone 4S that would download an uncached copy of a list of URLs. The URLs I selected were all images in jpeg or png format from my own web server and a variety of well known websites, including Apple, Amazon, Google Maps, and Flickr. I ran my app once on my WiFi network at home (connected to the internet via Comcast) and then again with WiFi disabled, using Verizon’s data network.

Finally, I copied the images off my iPhone using Xcode.

Results

Of the fourteen images I downloaded, nine had been altered by Verizon. The savings were significant: the images downloaded via WiFi totaled 2.1MB, while the same images fetched over Verizon totaled 796KB. That’s less than 50% of the original size, despite the fact that five of the images had been left untouched. Among the files left untouched was a 500x333px image from flickr and some of Google’s map tiles. Satellite tiles from Google Maps were not spared.

For the images that were modified by Verizon’s network optimization, the results were mixed. One photo of my cat showed an 80% reduction in file size with a minimal, but still discernible, effect on the quality of the image, while the quality of Atomic Object’s logo was very visibly sacrificed in the effort to bump it down from 27KB to 5.3KB. In another case, a product photo from Amazon suffered noticeably in quality for the sake of saving 2.7KB.

It’s worth noting that in all cases where Verizon applied re-compression, it also stripped out almost all image metadata.

Conclusion

Unlike Verizon claims, it’s clear that the loss of quality is not imperceptible, although it was very hard to tell with some images. To Verizon’s credit, the reduction in quality certainly pays off in the end with significantly less data to transfer, leading to faster load times. I believe this technique helps explain why I have observed that Verizon iPhones tend to load web pages faster than their direct AT&T counterparts, despite AT&T having a theoretically faster data network. Additionally, if Verizon is honest in their recording of data usage, it could also help you get more out of your metered data plan than you would with AT&T.

Despite the benefits, it’s still rather unnerving to know that your data is being mutated on its way to you. You might want to think twice about downloading new wallpaper over 3G. Or anything else where quality matters. But, thankfully, network optimization is only applied to HTTP traffic over port 80, so, for example, your email and encrypted web sites should be fine.