Visitor Monitoring Speed and Tenacity: The Apple iPad Outage | Keynote
Blog

Speed and Tenacity: The Apple iPad Outage

By Aaron Rudger | March 21, 2012

CATEGORIES: Web Performance

We’ve heard a lot recently about the importance of speed and performance when it comes to online retail. The New York Times highlighted research from Microsoft claiming that 250 milliseconds—a mere eye blink—could make the difference between a repeat visitor and a lost customer. And a popular infographic touts that Amazon would stand to lose $1.6 billion in sales per year from a 1 second web page delay. Our friends at Walmart.com have also shared some awesome research linking web performance to conversion.

These statistics are welcome news for the web performance community. But sometimes they don’t apply. With Apple, a lot of rules don’t apply.

Apple-ipad

This past weekend, Apple sold a record 3 million new iPad 3 tablets. That’s pretty phenomenal. Yet, it came on the back of a pretty bad outage only 10 days before.

Apple-store-scatter

On March 7, Apple announced the new iPad 3. For effectively the entire day, the Apple Store was unavailable. That meant no one could check out the new iPad, nor purchase iPhones, MacBooks or anything else.

To Apple’s credit, the Apple store normally runs very quickly—averaging well less than 2 seconds for totalUser Experience Time and less a second for Time to First Paint. (The Apple Store is a member of theKeynote Retail Performance Index, measured withKeynote Transaction Perspective.)

We’ve written previously about the concept of tenacity. A website visitor’s tolerance for errors, or delays, is a major factor when balancing the cost and benefit of building capacity and engineering performance into Web applications. While Apple’s fanatic customer base is an extreme, it illustrates the point that there’s a continuum of performance expectations for users.

Apple-store-trend

Your product/service is unique. And your customers are also unique. Keynote web load testing consultants dig into web analytics to model user behavior. They consider familiarity, tenacity, interaction speed and connection speed when developing virtual user profiles. It may be unrealistic for you to understand how different levels of performance impact your various customer types across all these variables. But if you can begin to understand them, you’ll be in a better position for setting ongoing performance goals and SLAs—especially around tolerances for outliers from your averages.

Back to Top