The Super Bowl is no different: it’s a day when the best ad agencies and largest companies in the US compete to one-up each other with funny (and expensive!) ads that get massive exposure – most of which drive a large volume of traffic back to their websites.

This year we saw advertisements from the following companies:

• Audi
• Best Buy
• Bridgestone Tires
• CareerBuilder
• Cars.com
• Chrysler
• Coca-Cola
• Dannon Yogurt
• Doritos
• E*TRADE
• General Motors
• Go Daddy
• History Channel
• M&Ms
• MetLife
• Sketches
• Toyota
• VW
• Walt Disney Pictures
• And more!

To see how they would perform, we set up 15 minute monitors from the following locations:

• Portland, OR
• Washington, DC
• San Francisco, CA
• Dallas, TX
• Chicago, IL

Overall all of the advertisers did fairly well: very few had major outages (with a couple exceptions). We did notice some sites slowed down just after their ads displayed. For example, CareerBuilder, as you can see from our monitoring results, slowed from a 3 second load time to a ~3.5 second load time during the big game:

(Click to enlarge)

Oddly enough, we noticed that E*TRADE performed great during the actual game, but stumbled hours beforehand:

(Click to enlarge)

We were curious about this slowdown earlier in the day, so we dug in to it and found that they had some Adobe Omniture tags that were timing out, as you can see from our “waterfall chart” that shows all the objects on the page:

Incidentally, we can’t find those files referenced on E*TRADE’s site as of Monday, February 6th, so perhaps they removed them due to the disruption the day before.

(Click to enlarge)

Probably the biggest trend we noticed was that many of the sites employed the use of video on their landing pages. Most performed well, but we did see a few sites, such as Coke’s Polar Bear ad, available at http://cokepolarbowl.com, crashed pretty hard during the game.

It got so bad that they eventually ended up simply redirecting the site to a video of their commercial on YouTube, which was able to handle the load just fine. Not exactly the kind of result you’d expect after spending $3M+ for an advertising spot!

Related articles:

Which commercials crashed websites, and other Super Bowl Tech challenges

GoDaddy Didn’t Crash(and other lessons on predicting doom)

Leading order fulfillment and supply chain management provider Zoyto truly believes this mantra too. The company uses Webmetrics Full-Service Load Testing services to test mission critical systems prior to releasing service upgrades, and their experiences have even taught us a few best practices when it comes to relying on load testing for increased sales. Find the top three tips they’ve taught us below.

Ensure data accuracy.
The key to good load testing is ensuring meaningful data that helps you identify ways to improve your system performance. Zyoto gathers significant amounts of performance-related data beyond what is generated by unit testing to ensure data accuracy. For example, the company recently experienced a real-life situation similar to their Webmetrics load tests. Because the test results were close to the actual outcomes, Zyoto salespersons knew how to address the real-life situation based on previous testing scenarios and, inevitably, kept their customers happy.

Know your unique testing vulnerabilities.
In order to prevent negative impacts on our customers when implementing new services or launching new products, you need to first and foremost find potential problems in a testing environment. Zyoto takes their testing methods a step further to understand what scenarios have a higher likelihood of serving as the company’s Achilles’ heel. Zyoto knows that potential challenges are more likely to appear when there is a peak load on the server. For example, during a recent test they discovered an issue that one of their software vendors needed to correct.

Test before a product / service rollout.
Zyoto continues to use Webmetrics to load test before major releases so their systems perform beyond the levels expected by our clients. In fact, Zoyto has seen significant performance improvements in its systems. Two recent load tests demonstrated a 59% increase in successful page views per minute and a 38% increase in the number of concurrent users that can be handled.

To learn more about best practices learned by Zyoto, find the full case study on Neustar Webmetrics.

Many shoppers hoping to get the best deals on the newest electronics will spend hours while at work buying for their friends and family.  The only thing more frustrating than being caught by your boss while shopping online is being faced with website that is slow or offline.

According to ComScore, online sales are expected to increase 15% in 2011.  Of this 15% increase, the average online shopper spends $186.33 online on Black Friday and through the Thanksgiving weekend.

Leading up to Black Friday and Cyber Monday online sales have been higher than the same time period last year. Chase Paymentech is reporting a 25% increase in volume.  As Black Friday signifies the beginning of the 2011 holiday buying season, US holiday ecommerce sales are expected to hit $32.6 billion for the year.

Cyber Mondays Past

In years past, some websites were unable to support the increase in traffic during the holiday weekend. Many sites came offline for a period of time during Cyber Monday due to the flood of visitors trying to get in on a sale.

During Cyber Monday of 2009, the Kohl’s website was unavailable to approximately five percent of the traffic to the site. This is five percent of sales Kohl’s did not see due to not being prepared for the traffic.

Kohl’s is not the only site to have issues due to traffic on Cyber Monday. Gap.com also faced downtime in 2008.  The Gap website was down for an hour and a half.

What does it mean for you when your site is down on Cyber Monday? It means loss of revenue and a tarnished brand. On what is now deemed the biggest online shopping day of the year, every minute an e-retailer site is down is loss of money.  One way to make up for downtime is to be reactive.

In 2009, the Lenovo site could not handle the drastic increase in traffic that caused the site to go down. To make amends to consumers who were unable to access the site, Lenovo extended their sale through Wednesday.

Cyber Mondays Future

Online sales are not expected to slow down in the near future. By 2014, the estimated amount spent in online sales is $250 billion dollars. Are you prepared for these customers?

The best way to make up for downtime is to be proactive. Load test your site to make sure you have the ability to handle the extra traffic during Cyber Monday.  Monitor your website to make sure that if there are any problems, you know about them right away.

Learn from the mistakes of Cyber Mondays past, and prepare for the Cyber Mondays future. Check out our load testing and web monitoring solutions to ensure the best uptime for your site through this holiday season.

This means your mobile website needs to be up to par for a complete customer experience that rivals a customer experience on a PC. We’ve outlined a few tips to help you prepare your mobile site for the future.

Know your customer’s mobile experience.

It is impossible to know what your customers truly experience on your mobile site without seeing for yourself firsthand. Real device monitoring allows you to identify, track and fix problems with your customers’ mobile experience in real-time. This allows you to make necessary improvements at a moment’s notice.

Look at the big picture.

An effective mobile monitoring strategy includes an analysis of trending data via detailed performance logs, graphs and reports. Mobile Web performance isn’t just about downtime—it’s about load time, uptime, and content management, among others. Make sure your mobile monitoring solution provides complete visibility into every function of your mobile site.

Evolve with the 21^st century. According to the Ericsson report, most of the mobile traffic growth will be driven by video. Mobile monitoring should incorporate more than the analysis of your mobile site to meet the high tech expectations of today’s consumer. Enlist a mobile monitoring solution that analyzes embedded videos, SMS, mobile streams and downloadable applications.

For more information on Webmetrics Mobile Monitoring solutions, visit http://www.webmetrics.com/monitoring-services/enterprise-monitoring/services/mobile-monitoring/overview

As helpful as we hope these guidelines are, we know sometimes our readers want to hear how load testing works from the end users themselves. We’ve compiled a few best practices that have come right out of the mouths of customers using our load testing services:

Test different types of traffic at different times of the day.

“We started by using a model of expected user behavior—consumers, manufacturers, and retailers—and testing features individually for each type of traffic. We see higher traffic at certain times of the day, so that was a factor too. We tested lots and lots of scenarios and finally, to be really sure, ran a ‘perfect storm’ scenario with traffic spikes from all our users.” – Dave Cumberland, VP of engineering, Shopatron

Test when you expect hikes in traffic.

“We had a good handle on the size and scope of the stress tests to run… It took a couple of days to build the scripts and subsequently each test iteration was completed within a single day. We were able to demonstrate that the platform could easily handle the expected load over the holiday season after just one iteration. We plan to execute additional iterations after significant code changes or in anticipation of events that will impact system load.” – Steve Romney, CTO, Shutl

Replicate the true user experience.
“First, we tested what it’s like when people log in [to our online auction site]. Then we replicated heavy bidding, with bids coming every 15 seconds. Finally, they tested the refresh process, where bidders get real-time updates. This allowed us to pinpoint what to expect during peak loads.” – Brian Ellis, Director of IT, Taylor and Martin

For more information on how we’ve helped our customers and what they have to say about our load testing services, take a look at our case studies here.

Do you have any best practices to share from your own load testing experiences?

As a result, companies need to make sure they can handle this forecasted boost in users that are planning to make purchases through mobile browsers and applications.

The first step to ensuring superior application performance is to test and analyze every aspect of your web-enabled business so that you can iron out potential problems before your customers notice them.

Application monitoring frequently—sometimes every minute—simulates defined web transactions, such as customer logins and purchase order fulfillment, to verify that web applications are available and performing 24/7. By evaluating load times, Flash processing and end-to-end transaction times, you’ll know how to improve your applications for customers around the world.

After all, Gartner’s findings are not exclusive to the U.S. The research firm predicts that by 2013, 80% of North American and European online sellers will expand into Brazil, Russia, India, Africa, Japan or China. This is hardly surprising with the recent growth if Internet usage in Africa and India. One of the top benefits of running a web-enabled business is the prospect of greater global expansion opportunities.

Here at Webmetrics, we let you decide which locations to use in your monitoring service. Our monitoring agents are located throughout North America, Asia, Africa, and Europe, giving you a global perspective on your local issues. We also help you select a subset of locations to use in reports to better determine your actual performance variation.

For more information on our application monitoring services, check out our data sheet here.

What are you doing to prepare for this mobile revolution? Tell us in the comments section below.

Before Cloud Computing and the advent of a true hourly computing model, this would have meant provisioning hundreds of servers in multiple data centers and ensuring the proper upstream circuits were in place to handle the traffic. This effort alone would have previously taken many weeks to turn around with all of the headaches of making sure the hardware is properly racked/stacked, memory is in stock, server builds working as expected, etc, etc.

However, as more and more hosting providers move to a Cloud computing model and infrastructure, for this project we were able to provision the large amount of infrastructure and aggregate throughput capability for this test in less than a week (including contracts, servers builds, testing, etc.). Utilization of cloud resources also helped us to dramatically reduce the total project cost across the board. These types of large, bandwidth-intensive tests have always been very expensive and unless you are testing for a very large roll-out, the cost of doing network throughput testing has never really stacked up too well against the value of the testing itself. With Cloud computing, this has changed dramatically. The servers involved in the test can be spun up for 1-2 hours, generate the required traffic and then be quickly taken offline. What would have previously cost hundreds of thousands of dollars in infrastructure, can now be done for a fraction of the cost. Due to the high amount of throughput required for this test, we needed to split the load across multiple data centers, backbones and providers. We were able to quickly work with many different cloud providers to drive the required network traffic from multiple locations throughout the world.

I’d like to share a couple of key statistics from the testing that we were able to achieve by harnessing all of these cloud resources and multiple networks:

• Transferred 20+ Terabytes of data within 60 minutes

• 10+ tests executed within 10 days.

• 750+ machines used to generate the load from 10+ data centers

• Peak throughput achieved was 43 Gbit/s.

• Processed 1,000+ transactions per second under peak load (see graph below)

• Multiple tests identified several network bottlenecks in the Customer’s infrastructure which were then researched and resolved prior to go-live.

At the end of the day, we were able to harness the power of the Cloud to quickly turn around a huge and complex test for our Customer on what would have previously been an impossible task in terms of scheduling and logistics.

What are your testing needs? Have a large project you need to turn around quickly? A complex website that requires real-browser testing? Whatever your requirements, our Professional Services team has the tools, experience and technology to make sure your site gets tested right the first time. Drop us a line and let us know how we can help you achieve your testing objectives!

Below we’ve outlined a few of his key points that will help you monitor your site so you can sleep well at night knowing your website is always up and running.

Monitor everything. You need to be alerted whenever a critical system process goes offline. This includes web servers, cache, servers, databases, etc. Understand that you may not know everything you should monitor until it breaks for the first time, but that only serves as lessons learned for additional metrics to monitor and make for a more failsafe website over time.

How you monitor matters. Whether the purpose of your website is to log in a user or add an item to a shopping cart, your customers depend on each click of the mouse to be fully functional. These interactions require monitoring that automates real browser interactions—like sending headers, setting cookies or executing JavaScript—and alerts you if something unexpected occurs. This provides a realistic view of the customer experience.

Monitor from multiple locations. To ensure your website is fully functional from every coordinate on the globe, you need to monitor your site from different geographic locations. There is no “magic number” for how many locations you should monitor from, but we recommend at least three. For example, if your company is based in the U.S., we suggest you monitor from two locations in the States and at least one in Europe or Asia.

Implement Real User Monitoring (RUM). In addition to monitoring your website uptime, you should also test the load times for your website visitors. Similar to Google Analytics, RUM involves placing a JavaScript beacon on your site. Every time a user visits your site, the beacon records the load time and you are alerted whenever this time goes above a specified threshold.

Establish tiered alerting. If a tree falls down in a forest, does it make a sound? The same thought process behind this popular riddle goes into a tiered alerting system. If the individual responsible for fixing a downtime issue doesn’t check his or her email at 3:00am regularly, with a simple email alert system he or she will not hear if/when your site goes offline. Services like PagerDuty ensure that alerts run up the chain of command with continual forms of alerts until someone responds via email, text message or phone call.

For more information on Ian’s response or to join in on the discussion, click here.

This huge jump in mobile web traffic means that your corporate website—the central hub of information, resources and communication for your customers—is now more important than ever on a mobile platform. Here at Webmetrics, we cannot stress enough the value of mobile monitoring so that your customers have a fast, seamless experience from any device.

The report also showed that the majority of mobile traffic still originates from small screen devices, reinforcing the value of monitoring from real mobile devices. Tools like Webmetrics mobile device monitoring provide a highly accurate view of the mobile application and website performance your users are experiencing on their devices—from iPhones to Androids.

The key takeaway to remember from the data in Walker Sands’s web traffic report is that it is important to know the in’s and out’s of your customers’ mobile experience. This will help you solidify your position as a leader in your mobile-enabled business.

For more information on mobile monitoring, check out our resources guide here.

• 215.1 million Americans used the Internet in July 2011
• The average user was online 25.3 days in the month, for a total of 34.8 hours, and viewed 3,029 pages
• 86% of Internet visitors viewed an online video in the U.S., for an average of 18.2 hours each

Businesses need to follow and understand Internet trends so they can best serve their customers and develop business objectives that drive their bottom line. This surge in Internet users means businesses need to prepare their corporate websites and applications to handle spikes in web traffic.

Website monitoring and load testing are two essential tools that help businesses keep up with these trends. Real-time external monitoring and full-service load testing help organizations gauge website and application performance at various levels of load from the end-user perspective. This is key to identifying and fixing problems before customers ever become aware of them.

Around this time last year, Internet startup retail delivery supplier Shutl anticipated a significant increase in business due to the upcoming holiday shopping season and planned service expansion for customers. Shutl used Webmetrics monitoring and load testing solutions to ensure its online services could support rapidly growing demand.

Today, Shutl has greater visibility into the success of its website and can address performance problems that directly impact current revenue and long-term business growth.

More information on how Shutl uses Webmetrics solutions can be found here.