Online sales during the 2012 holiday shopping season topped $4 billion just between Thanksgiving and Cyber Monday, with analysts predicting even bigger online profits in 2013.
What’s at stake: Billions of dollars in revenue during the most important shopping period of the year.
Adobe forecasts that Cyber Monday sales will increase by 15 percent year-over-year, Black Friday sales by 17 percent and Thanksgiving Day sales by 21 percent.
The Web’s importance as a retail channel is increasing. According to IBM’s most recent Digital Analytics Benchmark data, online sales were up 110 percent year over year in the first week of November, while mobile sales were up by 43 percent. Smartphones and tablets continue to be an increasingly crucial part of the picture. On Cyber Monday 2012, more than 18 percent of shoppers used a mobile device to visit a retail site, an increase of over 70 percent from 2011, while mobile sales themselves also increased over 96 percent.
All of this activity means that retailer’s websites are under significantly more stress during the holiday shopping period than usual. More visitors can translate into bigger profits, but if the website slows down or breaks down, the potential losses are also larger than they would ordinarily be. According to a report by the Ponemon Institute, one hour of downtime on Cyber Monday is estimated to represent a loss of almost $500,000 or about $8,000 per minute. And the estimated brand damage could average $3.4 million or more per hour of downtime.
Even if a site’s just a little slower than usual, any glitch or delay in web performance will result in lower sales as customers flock to faster sites.
There are number of steps that online stores should take to ensure they get the most out of the holiday season:
- Test for the worst case scenario. At Shopify, an e-commerce platform with over 75,000 active stores, 124 sales were processed per second during Cyber Monday 2012 -- with no down time. For 2013, it’s testing its platform to ensure fast page loads for all their stores. “We created something called Conan,” says Harley Finkelstein, Chief Platform Officer at Shopify. “It’s a simulator to stress test the entire platform that truly represents the potential amount of traffic that could happen. No matter what traffic we get we want to make sure nothing will take it down.” Just doubling the numbers from last year isn’t enough – Shopify wants the site to work even if traffic goes up 10 times.
- Boost everything. Once capacity has been tested, it’s time to scale up. In preparation for peak shopping, Shopify upgraded its master database servers, and increased its page caching layer by 400 percent, which increases cache hits by 20 percent and delivers nearly instant page loads. It also doubled its application servers, which run the pages for storefronts, so that Shopify could serve more clients at the same time. “Some payment gateways can be slower,” says Finkelstein. “So we moved it to the backend so that front end resources can do other things.”
- Know your applications. “With VMs [virtual machines] you can scale up fast, but the key is how you load the application into that and get it in cohesively,” says Kevin Conway, director of the global solutions management organization at Savvis. “There are still intricacies with each application. You have to understand the application to do that effectively and have the teams and the process in place to do that well.” Optimizing these applications, for something like page load speed, for example, requires a detailed look at how they run in the overall infrastructure. “How does it utilize memory?” says Conway. “It could be a bandwidth issue but usually it’s a component between internet connection and the servers.” One ecommerce platform Savvis worked with was making calls outside of the data center so that it can authenticate payments, adding 17 milliseconds to the transaction time. “Consumers end up having a terrible experience because of that latency,” Conway says.
- Dive into integration points. “We know how to scale up a website to support peak periods and we know what we need to do to expand out web tiers and application tiers,” says Conway, describing one client who is moving its e-commerce system in-house after years of outsourcing. The client is taking a smaller site and moving it onto an IBM platform hosted by Savvis with a systems integrator working on development. “The bigger concern is the back end system. We understand the mapping of those integration points between Savvis and the customer’s data center as well as a co-location that they have running with us. You have to test those integration points for what the future will look like.”
- Have a Plan B. Retail Web sites should be housed in more than one location. “A retailer can have clones of their Web site infrastructure in different data centers,” says Gerardo Dada, director of solutions marketing at Rackspace. “A retailer could have their entire store in the Dallas data center. A copy of that infrastructure could be built in Chicago. Then you can allow those data center clone infrastructures in the cloud to grow based on demand. The copy in Chicago could be a little brother -- functionally the same, but at a smaller scale. But you can grow your capacity in Chicago to serve what Dallas can’t.” For global retailers, traffic can be distributed based on which data centers are closest to customers. And if a site goes down, one of the clones can pop up to take its place.
- Get “bursting.” Cloud bursting, that is. Basically, when computing resources run out in one location, the additional workload can be “bursted” elsewhere depending on demand. “We offer monitoring tools to our customers to look at how traffic is going, performance, page load, server capacity and when it reaches a threshold, you can automatically scale,” says Dada. “So for example if page load goes over 2 seconds, you can spin up more servers to dynamically grow based on demand. As people leave work at 5 and demand increases, you can grow infrastructure by the hour.”
- Above all, a retail website should work, without delay or fail. Getting all the pieces in place both on the front and back end ahead of time is key. “The key driver is the customer experience,” says Conway. “That’s at the top of the mind with every executive you talk to. You have to dive down into how you deliver that experience.
Black Friday gets earlier every year. For brick and mortar stores, that means people are lining up in the dead of night to get an even earlier start on seasonal deals. From an online perspective, it means that retailers have to be ready for shoppers even sooner than in past years.
“If retailers haven’t already looked at their environments and tested them for security and performance, it’s probably too late,” says Amin Shahidi, VP, Consumer and Retail Industries for HP Enterprise Services.
Retailers who have their Black Friday strategy perfected don’t start the day before, or the week before, or even the month before. They start right after the previous year’s Black Friday, and have a good sense of when shopping spikes will occur throughout the year to come.
“This is a constant process. It isn’t something that happens once a year,” says Saif Rivers, Retail Client Principal for HP Enterprise Services. “Though the event may be once a year, immediately after we go into a process of understanding what changes were made, what worked and what didn’t, so we go into our next planning cycle with what we’ve learned from the last year.”
As always, it’s all about knowing the customer. When it comes to e-commerce, it’s especially about knowing what makes the customer mad.
According to a survey by Akamai, almost half of users want a site to load in two seconds or less, and will abandon a site after three seconds. Close to 80 percent of shoppers who have trouble with web performance say they wouldn’t go back to the site again.
“The cloud has enabled retailers to scale to the point where there’s no excuse to be down over the holiday season,” says Gerardo Dada, director of solutions marketing at Rackspace.
Another area retailers fall prey to when it comes to disappointing consumers is in the synchronization between their physical stores and their online stores. “You need to make sure that stock in store can fulfill an online process,” says Rivers. “So that someone can buy online and pick up in store.”
And if a retailer is running deals, it needs to make sure that they’re in place across both environments. “If promotions are in place online and not in the store, you’ll have a lot of people coming into the store and complaining. You’ve pushed the sale into the hands of your competitors,” Rivers says.
- IBM’s Black Friday Report 2012
- IBM’s Cyber Monday Report 2012
- 2013 eCommerce Cyber Crime Report
- Adobe’s 2013 Online Shopping Digital Index
- As Shopping Season Expands, Is Black Friday Dead?
- ComScore’s Cyber Monday 2012 Report
Amy Lee is a business and technology reporter for CruxialCIO. She has written on technology for the Huffington Post and is a 2010 graduate of Yale University in English and writing.