Hard-to-find Web glitches cost companies in lost or disconnected customers

In a startling look at what hard-to-find, intermittent problems within the Web infrastructure can cost a company in terms of Web performance and customer satisfaction, Mercury Interactive has released a report finding that Web sites, overall, are performing on par only 63% of the time.

The report, titled Challenging Industry Perceptions of Web Performance, compiles key statistics of Web performance and availability across 15 global industries, and challenges popular rule-of-thumb approaches to measuring Web performance and its affect on users. It also shows that for IT managers setting 99.99% uptime as a target for Web site availability, the industry has a way to go to reach that goal.

Graeme Allcock, sales director for Mercury Interactive in SA, says the report highlights data collected over the past year-and-a-half by Mercury Interactive`s application performance management experts, who have monitored thousands of e-business transactions at hundreds of large Web sites worldwide with ActiveWatch. ActiveWatch is Mercury Interactive`s hosted Web performance monitoring and diagnostic service powered by Topaz, a modular platform for application performance monitoring and diagnostics. The entire report can be seen online at http://www.mercuryinteractive.com.

"Many Web sites support business functions that require continuous or predictable availability, all the way to the end-user," says Bill Gassman, senior research analyst with Gartner. "Without proper monitoring, it is hard to know when users experience intermittent problems, or the business impact of those problems. The impact can result in lost sales and a tarnished reputation, so large returns on Web site monitoring investments are common."

One of the most interesting facts to surface in the report is that approximately 70% of all site unavailability cases are generated by intermittent problems - companies are rarely aware of them - and it is almost impossible to recreate and pinpoint these types of glitches without automated monitoring and diagnostic tools. It becomes clear that, to find these problems, companies must monitor entire business transactions all the way to the end-user, and not just look at page load times. Report findings show that popular measurement thresholds, such as the "eight second rule" - that users abandon sites if a page doesn`t load properly in under eight seconds - is not an accurate way to measure real performance, as it depends on the type of transaction being measured and what the user is trying to accomplish.

For example, the ActiveWatch data shows the following transaction times:

• .         Average home page download time: four seconds

• .         Average login time: 12 seconds

• .         Average search function time: 10 seconds

"The ActiveWatch data illustrates that there is room for variance in performance thresholds, and that the eight second rule approach is not the most effective way to measure performance," Allcock says. "Rather, organisations need to set performance thresholds based on their users` business needs and expectations. If users feel the data is critical, they may not mind waiting longer than eight seconds for a transaction to be completed. Conversely, for pages with static text, even a few seconds may be too long.

"What we`ve learned from monitoring thousands of transactions is that without an automated monitoring and diagnostic solution, companies can fail to spot major problems," he continues. "ActiveWatch and Topaz give IT the ability to find the root cause of hard-to-find performance problems. These are the problems that, left unchecked, can end up costing companies in upset or lost customers, and lost revenues. It`s not enough to monitor how fast or slow a page loads. You have to monitor the entire end user experience, and get to the root cause of a problem before it affects users, before the transaction fails - and the customer is lost forever."

Additional ActiveWatch findings show that, while major Web site crashes, the Code Red worm, DOS attacks and other high profile problems get the majority of public attention, companies are wise to sweat the little things that hinder Web performance, and eventually affect their users` experience. This includes unreliable, unavailable or slow Web applications; changes made to the infrastructure that directly affect end users; quickly knowing which departments and systems can resolve a problem; and what specific locations are impacted.

Other intermittent problems can substantially raise the level of difficulty in finding the root cause. Additional findings brought to light as a result of compiling the ActiveWatch results include:

• .         Availability in North America is higher, but not high enough - Overall Web availability for North America is 96%; worldwide availability averages 92.5%. Many IT managers set a goal of 99.99% availability. The report shows there is still a lot of work to do to make that a reality.

• .         Poor customer service - For nearly 75 % of all Web sites, one in 10 customers has a bad experience due to performance problems.

• .         Servers are the frequent culprits - Internal application component and Web servers are much more frequently to blame - up to 75 % of the time - for poor performance than are problems with the network outside the firewall.

• .         High thresholds can mislead - Companies may set performance thresholds at eight seconds for a page to load. If the page normally loads in one second, performance would have to degrade 800% before a warning is sent. This can cause IT to miss an opportunity to isolate and repair a problem that may have serious consequences later.