(WEB HOST INDUSTRY REVIEW) — Strangeloop (www.strangeloopnetworks.com) recently launched its new Site Optimizer solution, which shaves valuable seconds from page load times by approaching performance from the perspective of the browser.
With research suggesting that Internet users become impatient after waiting just two seconds for a page to load, surprisingly, the average Fortune 500 website still takes seven seconds to load. Only one of the 500 sites actually meets the two-second standard for page load time, despite the huge investment these companies have likely made in application delivery controllers, content delivery networks, and in-house development projects.
Launched in June, Strangeloop’s Site Optimizer analyzes every page of a site from the browser’s perspective and optimizes each page so it is delivered most efficiently to each browser type. It also delivers powerful Web acceleration features that work together to provide roundtrip reduction, rapid rendering, and dynamic browser caching.
The resulting site acceleration, in turn, is a proven way to increase ecommerce conversion, cart size, revenue, and search engine ranking – even for sites currently using ADCs and CDNs.
Strangeloop was founded in 2006, by a management team from a diverse array of Web environments, who all dealt with the problem of delivering robust, feature-rich sites without sacrificing the site’s speed.
The term “strange loop” was coined by Pulitzer prize-winning author and professor of cognitive science Douglas Hofstadter. Interested in the nature of consciousness and identity, Hofstadter developed the term while arguing that the phenomenon of self-awareness is best explained using a model based on symbols and self-referential “loops” that, as they accumulate experience, create high-level consciousness. Hofstadter argues that an intelligent system must have a self-referencing feedback loop, learn from that feedback, and adapt accordingly.
Strangeloop respectfully borrowed the term and the concept in its effort to create a solution that bypasses application code, and creates an intelligent, self-referencing device that can learn and adapt in real time.
In an email interview with the WHIR, Strangeloop president Joshua Bixby explains how the company’s Site Optimizer, inspired by Hofstadter, works in conjunction with other acceleration solutions to compound performance optimization.
WHIR: The impact of site downtime is obvious and well documented, but how much does page load time cost companies?
Joshua Bixby: To answer this question at a high level, I’ll cite an Aberdeen Group study from last year that states that issues with application performance – referring to web-based applications – are affecting overall business revenues by up to 9 percent.
To answer this question in a bit more detail, you need to approach it in two ways:
First, there’s the immediate cost in terms of lost revenue from a single transaction. Our recent studies of ecommerce sites show, on average, a 10 percent drop off in conversion for every second of delay. This relationship appears to be linear (see attachment: a recent graph that we are presenting at Velocity), which shows the relationship between performance of a landing page (i.e. the first page someone hits when they come to the site) and the likelihood that they will convert (i.e. buy, click, sign-up, etc.).
Not only do slower pages lead to lower conversion rates, they also result in smaller cart sizes. We know this because we know that optimizing performance has been proven to increase cart size, on average, by 5 to 6 percent. Based on these conversion and cart size stats, this means that if your site typically earns $100,000 a day, that one second of performance loss could cost you $2.5 million in sales.
Secondly, there’s the long-term cost of poor performance. Nine out of 10 visitors who have an unsatisfactory experience with a website are unlikely to return. Of these visitors, three will go on to tell others about their poor impression of the site. It’s impossible to assess the magnitude of lost site traffic and lost revenues, but we can deduce that, for a site that consistently performs poorly, these losses must be significant.
Approaching costs from yet another perspective, it’s interesting to note that, in 2010, companies are expected to spend a total of $4.2 billion on CDNs and network devices such as ADCs. These are helpful solutions, but they don’t solve the entire performance problem.
For more statistics about site performance and metrics, I’ve summed up a lot of data in this recent post on my blog.
WHIR: How does the Strangeloop solution deal with modern websites, which often include third-party services such as ad networks, widgets, etc.?
JB: Third party content is a serious issue with modern websites. Steve Souders has started a great project called P3PC, which is gathering crowdsourced data to analyze the performance of third-party content in real time with real users. So far, he’s found that Digg and Google AdSense have a big impact on how a page performs, while Facebook and Google Analytics keep pages fairly nimble. The idea with P3PC, as I understand it, is to identify which third-party services are worth evangelizing, in the hopes of establishing best practices across the board.
In the meantime, we’ve developed features in the Strangeloop Site Optimizer that help the delay imposed by third-party services by doing the following:
- Changing the order in which pages render content. Site Optimizer delays third-party roundtrips until after the page has loaded. (Obviously you can’t do this for all third-party roundtrips, so the research that Steve is doing is very important.)
- Caching and consolidating third-party trips whenever possible.
We can also work directly with vendors who create the third-party content. By implementing Site Optimizer in their datacenters, we can dramatically speed up how their content is delivered.
WHIR: One of the problems in monitoring the end-user experience is that individuals have different browsers, different connection speeds, and are in different locations. Can web acceleration factor in all those variables?
JB: The Strangeloop Site Optimizer does an extremely effective job of tackling performance from the browser perspective. First, Site Optimizer arranges each page’s resources optimally for each browser type, without affecting the functionality of the page. Site Optimizer also takes advantage of two valuable time periods: time when the server is thinking and time when the user is thinking.
While the server is thinking, Site Optimizer does two things. It offloads servers to allow faster web page generation, and it enables the web client to download each page’s resources before the server has generated the HTML. As a result, it dramatically improves the time it takes a browser to render a web page.
As for user think time, we consider that the time a user spends viewing a web page is valuable time that could be spent delivering resources to the user’s browser to enable subsequent pages to load more quickly, so we designed Site Optimizer to track and analyze users’ behavior patterns on a site. Using this data, Site Optimizer can predict which pages visitors are most likely to want to see next, based on the page they are currently on. Site Optimizer automatically pushes the resources – images, HTML, Javascript, etc. – for these pages to the visitor’s browser to have waiting on standby before the visitor clicks to see the page. The result is pages that load almost instantaneously.
WHIR: How much effort do Strangeloop customers have to put into the acceleration process? Does it involve re-working websites?
JB: The Strangeloop Site Optimizer automates the acceleration process. When it comes to our cloud-based service, our motto is “No installation. No code. No network changes. No problem.” It installs with, almost literally, a flip of a switch. Our hardware and virtual appliances do need some initial customer setup – with our assistance – up front, but this work is minimal and does not require hardware, server or network changes. No matter how Site Optimizer is delivered, it is designed to be implemented easily and yield immediate returns.
