Page Load Testing: Embedding With DocStoc and Scribd

As I work in search engine marketing – I frequently test my sites’ page load times just to make sure I’m getting my content to visitors as quickly as possible. I don’t spend a lot of time worrying about my blog – until I posted an embedded pdf on a post recently and noticed a significant performance drop off. I ran this site through the great tool at Webpagetest.org. I was appalled to see 10 second+ load times!!!1 I dug into the numbers and saw that my scribd embed was hogging bandwidth like a sonofabitch. I knew there are basically two players in this field – scribd and docstoc so i figured I would test page load time with each. The results are pretty interesting:

Page Browser Load Time Fully Loaded Req Bytes in
DocStoc Chrome ✓ 0.616s 4.324s 12 258 KB
DocStoc Firefox 1.295s 3.681s 12 258 KB
DocStoc IE8 1.348s 3.900s 12 258 KB
Scribd Firefox 7.800s 8.556s 49 557 KB
Scribd Chrome 8.378s 9.198s 57 552 KB
Scribd IE8 15.269s 15.755s 63 742 KB

DocStoc loads a flash object – while scribd loads up a remote page via an iframe. Intuitively one might expect flash to perform worse – but in this case the biggest culprit is the sheer volume of third party scripts scribd is embedding. Depending on browser between 49-63 request compared to DocStoc’s consistent 12. Additionally scribd serves up over twice the total page weight as does DocStoc (257kb vs 552-743kb). Scribd clearly isn’t concerned enuogh about user experience with their embedded documents.

Some notes about the test – first off its clearly unscientific – but the difference in load time was such that further tests seem pretty pointless. I used the default embed code from each site – and uploaded the same simple text document to each (a simple robots.txt file). Each page was identical and loaded no other external resources. The pages can be seen here and here.

A few items of interest – scribd is including twitter, facebook and google+ external javascripts which is probably not as noticeable on many sites – but still a very heavy bandwidth decision. The one item that stands apart though is scribd’s decision to include a quantcast tracking code (http://edge.quantserve.com/quant.js). Looks like a bid to inflate their quantcast numbers – judging by their traffic there.