- New Customer Discount: If your company first purchased Akamai services after January 1, 2013, you're eligible to receive a $300 discount on the Full Conference pass. Enter code: 50NEW2013
- Edge Alumni Discount: If you've attended an Akamai conference before now you're eligible to receive US$300 off your conference seat. Enter code: ALUM2013
- Team Discount: Visit Akamai Edge 2012 With Your Team of 3 or more and save 50% per person. Enter code: GROUP2013
Get In Touch
Recently in Retail & Commerce Category
The holiday season is already creeping up and by far will be the most vital online shopping period of the year for retailers. Thanksgiving week will no doubt once again present one of the largest online shopping weeks of the year. Now, more than ever, time is literally money when it comes to Web performance.
As online retailers face another big holiday shopping season, they have to make sure they are ready for everything and anything that will hit their site once the critical holiday season begins. Even a minute of downtime can cost thousands and thousands of dollars and can damage the bottom line and the brand for years to come.
Supporting the majority of today's leading global retailers, Akamai has the expertise and technology to help retailers simplify the process of preparing their e-commerce site for Black Friday as well as helping them to deliver on every big shopping day before and after with situational performance - so that any customer, anywhere can make a purchase as soon as inspiration strikes.
Starting this week, Akamai will be launching a series of "Crush The Rush" Holiday Readiness Webinars to help retailers understand how to properly prepare their site for the biggest shopping season to-date. Participants will learn first hand how leading retailers have deployed Akamai Solutions specially tuned for holiday traffic, ensuring their sites are scalable and reliable as customers flood their online store.
The "Crush the Rush" series consists of three Webinars focused on different topics relevant to preparing for the rush of the online holiday season. The first one will be held on July 24th on Situational Performance. Participants will come away with clear understanding of why looking at performance from an end-user perspective is key for those who wish to "Crush the Rush" this holiday season. Our speaker, Lorenz Jakober, will be discussing what is Real-User Monitoring (RUM), synthetic vs. RUM testing and the pros and cons for each, why optimizing beyond average load times matters and how performance optimization helps meet rising consumer expectations.
Following Situational Performance, we will also be discussing how Akamai Services & Support engage with the industry's leading brands as well as Web Security and Risk Mitigation throughout the Webinar series. Our speakers will be sharing recaps and best practices that retailers should be aware of for this holiday season. Follow us here to find out more details on the "Crush the Rush" Holiday Readiness Webinar Series. We want to help you beat the rush to the holiday rush.
Two weeks ago, a large number of eCommerce professionals converged in Chicago for IRCE 2013. I was one of the presenters to talk about "Building m-commerce: Different approaches, different outcomes."
There were plenty of instances throughout the conference where the growing mobile traffic and revenue numbers highlighted the importance of delivering fast, quality mobile experiences to consumers. As most of us know - fast, quality mobile experiences are better for business.
Yet as most of us also know, delivering fast, quality mobile experiences isn't exactly easy. The challenges associated with mobile performance are well documented (browsers, network, devices, etc.). For those who are interested in diving a little deeper into the topic, I recommend watching Ilya Grigorik's Google I/O talk - Mobile Performance from the Radio Up: Battery, Latency and Bandwidth Optimization.
The fact remains shoppers expect fast, quality mobile Web and app experiences and they generally don't know or care about the technological challenges associated with delivering them. In addition, as we start to think about the latest techniques to engage mobile users such as Responsive Web Design (RWD) these challenges become even greater. Responsive Web Design is a Web development approach that suggests Web pages should respond to the context in which they're loaded (primarily screen size) and change their user interface accordingly.
So what does delivering large, complex pages to mobile devices mean from an end-user's perspective? Below is a snapshot of the experience of an end-user visiting a US retailer's RWD site's home page on a variety of different devices/networks. The conclusion is obvious. The delivery of a relatively small 700KB site to a mobile device, over wireless networks, has resulted in serious performance shortcomings.
The first step to deliver fast, quality RWD sites is to focus on the actual page and the associated objects delivered to the end-user. As Web performance optimization guru Steve Souder likes to point out: "80-90 percent of end-user response time is spent on the frontend. Start there."
- Reducing the number of requests
- Reducing the number of bytes
- Accelerating rendering
For a more detailed view of how to actually reduce the number of requests & bytes and accelerate rendering download Akamai's Front-End Optimization primer.
Independent of your approach to engaging mobile users, it is always worth to remember the following:
- Deliver consistent, fast, quality web and application experience
- Adopt your customers' perspective
- Optimize for mobile first
Last year's Click Frenzy online sale in Australia, modeled after the hugely successful Black Friday sale in the US, attracted a huge amount of media attention for all the wrong reasons. Within minutes of the sale going live on November 20 last year, the website buckled under the strain of unanticipated traffic volumes. A number of consumers, excited at the prospect of grabbing a great online bargain, were left empty handed and disappointed. The media's reaction was swift, and merciless.
Fast-forward six months and the latest Click Frenzy sale was completed without any technical issues at all. Click Frenzy needed an industry leading solution, and as such enlisted Akamai.
Following their initial sale and the subsequent technical issues they encountered, Click Frenzy wanted to understand how they could handle the sudden bursts in traffic volumes that characterizes their online business model. For Click Frenzy, it was absolutely imperative that their next big sale performed seamlessly to restore consumer and retailer confidence. Further technical issues had to be avoided at all costs.
We approached Click Frenzy and explained how Akamai's intelligent platform - specifically Akamai's AQUA & KONA solutions - enables traffic to be securely offloaded to distributed computing resources, thereby alleviating the burden on the their data centre. This model allows businesses, such as Click Frenzy, to manage traffic spikes that would be unheard of for most conventional online retailers.
Following Click Frenzy's 24-hour sale last month, the Akamai platform managed 169 million requests, with a peak of 29,722 requests per second. Total traffic volume exceeded 3TB. This article, which features an interview I did following the sale, provides a great overview of some of the positive outcomes Click Frenzy has enjoyed since its rollout of the Akamai intelligent platform.
Although the Click Frenzy site itself performed flawlessly, some technical issues were still evident on the part of some retailers involved in the promotion since their websites - not utilizing Akamai's platform - were unable to cope with the traffic being directed at them by Click Frenzy.
For retailers, having too many customers can be a nice problem to have, but customers can be unforgiving if they are running into the same issues time and again. This is particularly true for online where a customer can 'enter' another store with a simple click of a mouse.
A robust technology platform that enables customers to access the information they need and purchase in an efficient manner is absolutely essential in today's super-competitive online retail environment. And retailers don't want to be bogged down by technology. They just want it to work, so they can focus on their core business and what they know best - retailing. For our friends at Click Frenzy, that's exactly what they're doing now and we look forward to many more successful online 'frenzies', and maybe picking up a bargain or two ourselves along the way.
Ian Teague is a regional sales manager at Akamai.
Akamai has been proudly partnering with Best Buy on eCommerce holiday preparations for years. While we work with many eCommerce IT organizations, Best Buy is notably one of the best - they are forward thinking and innovative, technically savvy about performance and scalability, operationally excellent with extensive monitoring, and have an excellent partnership with their business team.
In today's blog, I sat down with Andrew Tsai, the Akamai Engagement Manager for Best Buy, to talk about what makes the Best Buy and Akamai relationship so successful, and why Best Buy has an industry-leading holiday readiness model to emulate. Here's a photo of Andrew - we managed to catch him in one of the rare moments he's not eating a Chipotle burrito (or two).
[Lelah] I know you love working with the Best Buy team. How is Best Buy more sophisticated, in your estimation, than other retailers you've worked with?
[Andrew] They understand the benefits of availability, performance, and supportability and how working with Akamai helps accomplish those goals. When their business team wants bigger events, richer and more dynamic content say around holiday, RewardZone, or a Super Bowl commercial, it's executed with scale in mind.
We are part of the extensive collaboration that exists between their marketing and IT teams. As potential solutions are discussed, compromises are made if needed. In addition to scale, we do things in a way that we can support and track, so we can react quickly and apply lessons learned to the next event. Another differentiator is we have an enterprise type relationship with them and they have developed their own internal processes to deal with priorities and chargebacks as we work with many groups within Best Buy.
Finally, their operations team is excellent and among the best in class we've seen. They have extensive monitoring both at their origin and at Akamai. Both sides receive each other's alerts, so there is a huge collaboration between us so that either side can respond or escalate an issue quickly. Internally, they have great monitoring to keep third parties in check, trace individual user sessions, and have a very good understanding of their daily traffic patterns. They live in our Luna Control Portal, look at traffic and usage reports constantly, and have been known to recommend a few portal enhancements or two. In fact, before we even notified them, they once identified an Akamai caching change was almost complete because they saw their hits to origin dramatically decrease in real time.
From my perspective, the Best Buy team understands Akamai and the ROI of leveraging the Akamai platform. They always ask Akamai first. It's the difference between being a strategic partner and a vendor.
[Lelah] When does Best Buy begin holiday readiness planning? And what are their objectives for the season?
[Andrew] In all seriousness, it begins as early as Q1. Early in the year, we meet to recap the previous year, brainstorm ideas even crazy ones, and plan out the roadmap for the upcoming year. Obviously things come up, but it really helps us plan out the year so we have an idea of their initiatives, understand each other's team structure, and determine how Akamai's products and solutions fit in to their initiatives or not. The real heads down work starts roughly in August and they have a dedicated holiday readiness team who is responsible to make sure we are ready.
Revenue goals are of course primary, but they really look at offload and performance from an operations perspective.
[Lelah] In the architecture planning, how do you work with them to evaluate their architecture and their key risks?
Best Buy architects with Akamai in mind. We have weekly calls and we have quarterly reviews on site. If appropriate, we have access to internal sites and applications and share documentation whenever possible. We're not afraid to bounce ideas or innovate something new and often engage in quick proof of concepts so we can design with best practices and offload in mind. They are very smart with lots of cool ideas about how to use the Akamai platform. In fact, they were one of the integral customers that pioneered Akamai's SPA [Shopper Prioritization Application] product. Our account team loves it as we have a chance to architect and think outside the box.
[Lelah] Can you give me an example of how they use Akamai to scale?
[Andrew] Probably the best examples are their holiday initiatives for which we are key part. For their home page and doorbuster sales, they came up with static pages and used NetStorage to ensure that traffic was 100% offloaded to Akamai. Each of these pages was also designed with all types of devices in mind - desktop, tablet, iPhone/smart phones, and basic devices as well - and each of these device-specific pages is stored in NetStorage. We handled the mobile device detection and vanity redirects to the device-specific pages at the Akamai Edge. Time match logic was built in to automatically flip the logic during holiday peak times and revert it back after Cyber Monday. We also ran through all the purge scenarios in case of emergency changes. For pages we couldn't cache, we used ESI fragments and tweaked the automated purge schedule to minimize origin bandwidth. Key PDP pages and SKUs were also analyzed and architected in a way using GTM to go to different origins to achieve maximum offload.
This was the home page delivered from origin at 11:59 PM, right before the holiday sale kicked in:
At midnight on Thanksgiving, the Akamai configuration logic kicked in to serve this page automatically from NetStorage:
This happened for the mobile device pages too:
The result is 100% offload at peak!
[Lelah] Wow - that's impressive. Are there any other cool solutions that we've partnered with Best Buy on this holiday?
We've been working on a cutting-edge project for the last year that allows them to be very dynamic and scale on-demand with their traffic, be more agile with releases and maximize offload for their browse functionality. Unfortunately I can't share the details as its confidential to Best Buy, but Akamai has implemented some pretty sophisticated routing, caching, and failover logic that are critical to the project.
Recently, we've implemented a Download Manager for online games and their digital library that allows customers to optimize large software downloads purchased directly from bestbuy.com.
They also make great use of SPA [Shopper Prioritization Application] whether it's for planned maintenance or unforeseen events. They're smart enough to know that having a contingency plan is important, and they make sure the experience is as user friendly as possible.
[Lelah] In a previous blog post I talked with Paul about the criticality of load testing. How does Best Buy load test for the holidays?
[Andrew] It's part of their holiday readiness team and initiative that they have every year. Best Buy has been running load tests every few weeks since August. With Premium Support, our team is fully aligned to Best Buy and we are very much engaged to help them work through issues whether it's log analysis or pinpointing where certain bottlenecks may be.
[Lelah] This past Thanksgiving weekend, how was the Akamai team involved in monitoring?
[Andrew] In preparation, we put together extensive documentation on our internal wiki for our aligned Premium Support team so collaboratively we were well aware of their holiday projects and key URLs. We looked at past traffic peaks and since we were part of their holiday planning, we knew roughly when the traffic was going to hit us. We had our usual Akamai and origin alerts set up, but tuned for holiday to eliminate false positives. Bi-daily touch points were also a good way of keeping the Akamai and Best Buy teams in sync as we constantly monitored error rates and offload. Overall, things went smoothly and we were able to squeeze in some turkey/family/ football time, and I think that's a direct result of our close collaboration and proactive planning throughout the year.
[Lelah] I'm not surprised to hear things went so well. Thanks for sharing all of this great info on Best Buy's holiday preparations. You guys are definitely working on some cool projects. I now owe you a burrito (or two).
There are so many numbers and figures circulating since the key online shopping day: Thanksgiving, Black Friday, and Cyber Monday. But what exactly is the data telling us?
- Online Shopping Trends: when consumers come online to begin shopping for the holiday season, including key dates other than those during Thanksgiving weekend.
- Shopping Habits: whether the online consumers are coming online to browse, research or make a purchase or as part of their omni-channel shopping behaviors.
- Key Browsing Hours: some of the page view peaks, as well as the shopping data peaks discussed in my Cyber Monday post, provide insight from where consumers are likely going online: work vs. home.
Akamai Online Shopping Trends For November:
I want to hear from you. What patterns did your brand see? How will you approach 2013 differently? Which metric is the most important to your brand during this holiday season?
Please leave a comment, tweet @Akamai using #AkamaiHoliday, or to me directly @lyss3b.
Elyssa Duboys is a Senior Marketing Specialist at Akamai
Monday marked Cyber Monday in the UK; a day projected to be the largest online shopping day pre-Christmas, and second only to Boxing Day in all of 2012. Visa projected up to 6.8 million transactions on retail sites yesterday, spending up to 465 million GBP. That figure puts online sales up 21 per cent on the equivalent day last year. Experian projected the UK will make 115 million visits to online retailers, 36% more than last year's "Cyber Monday" which saw 84.6 million visits. The final data on growth has not yet been released, but early reports from retailers such as Marks and Spencer suggest that Cyber Monday delivered on the expectations.
Akamai works with dozens of retailers in the UK, and hundreds globally, and tracks the traffic, in page views per minute, across all retail sites in our Retail Net Usage Index. To look at the potential surge in traffic this Cyber Monday, we broke out UK traffic coming on all retail sites globally. Our data suggests that while Cyber Monday was a strong day, perhaps there is more hype in the size of the day than reality. In fact there are a handful of other retailers, like Sebastian James of Dixons, that are also skeptical of Cyber Monday's allure: "I don't think Cyber Monday has the same power it may have had. It dates back to a time when you had to wait two weeks for a delivery, but now we do 24 hour delivery and will offer reserve and collect up until Christmas Eve."
The chart below looks at the daily peaks and averages of UK retail traffic from the start of November through to Cyber Monday. While there is a marginal trend of growth in the Sunday and Monday peaks each week (shoppers typically research purchases on Sundays and buy on Mondays), the peak on Cyber Monday is only 6.5% higher than November 11th, during the mid-season sales. The averages have shifted slightly upwards as well, but outside of the mid-season sales, the trend notably begins on the US Black Friday, rather than Cyber Monday. Does this data suggest that UK shoppers are shopping earlier, and more globally? If so, there may be an opportunity for UK retailers to capture more of the local shopper's pounds if sales were to begin earlier in November.
Looking at the traffic changes under different lenses lends some additional interesting insights into shopping behavior:
- Saturday Belongs to the High Street: Saturdays drive the lowest amount of traffic, repeatedly, week after week. This is an indicator that shoppers are enjoying the experience of shopping on the high street on the weekends, rather than online retail sites.
- Sunday Research Day: Shoppers first come online in numbers on Sunday night. This is largely research-driven traffic as shoppers investigate the deals they are most interested in preparation for Cyber Monday. This is true not just during Cyber Monday weekend, but most weeks of the year.
- Cyber Monday Peaks in the Evening: On Cyber Monday, shoppers first start to come on in numbers at 7:00 AM, over their morning coffee, and continue coming online in numbers throughout the early workday, peaking at the lunch hour. In the afternoon, they get back to work, and then head back online again after dinner, ultimately peaking to its highest point at 8:00 PM.
We'll wrap up with the highlights on the all the growth reports as they come in. We'd also like to hear your stories. How did your site fare this Cyber Monday? Did your site see new peaks, similar to Marks and Spencer?
Now that Cyber Monday has passed, we've left the most daunting of the 2012 traffic peaks behind us. US retailers were largely stable and available this year, with few catastrophic site failures. In general, retailers are much more mature with their readiness planning, starting earlier, and conducting load tests and simulations right before peak.
For those of you tracking our Net Retail Traffic Index this past weekend, you may have noticed an odd spike of 11.5 million page views on the Sunday before Thanksgiving. A few savvy observers asked the very logical question - what's that spike? The biggest online retailers all engaged in "best case" load testing this past weekend.
Load testing is our number one recommendation for being prepared - you have to know your vulnerabilities. While this may seem obvious to many of you, it is surprising to hear the number of retailers who still attempt to squeak by without load testing before the holidays or peak events.
An Interview with Akamai's Load Testing Expert
Who better to speak with about Load Testing then Akamai's Load Testing expert, Paul Korenevsky. Fortunately Paul spoke with me last week, before he spent the entire weekend sleepless, onsite with his customer, watching the traffic and orders fly in. This is Paul (looking a lot more awake than he is this week).
[Lelah] Paul thanks for talking with me today as I know it's a particularly crazy time of year. Can you share what you do for Akamai?
[Paul] My primary responsibility is an Engagement Manager, which makes me the owner of the professional services team for our premium clients. My secondary responsibility is a SME [subject matter expert] for load testing best practices. We have defined such practices for load testing providers SOASTA and Keynote, and I am working to optimize those practices and expand the list of providers.
[Lelah] When a retailer runs load test, what are their primary goals?
[Paul] There are two goals. First they are trying to identify potential bottlenecks early. They will stress tests their obvious bottlenecks to understand how they will perform under load during their holiday peaks, and to tune where possible. Secondly they want to evaluate the performance degradation under load. The reality is that most retailers historically only got through to #1 but more and more they are starting to get to #2.
[Lelah] Which load testing providers does Akamai most commonly work with? How are they optimized to work with the Akamai cloud?
Most of the larger ecommerce companies offload not just static content but application processing onto Akamai. If they load test direct to origin, without Akamai, the traffic will not represent the load under real conditions.
Akamai initially didn't allow our customers to conduct load testing on our platform, without express consent of the Akamai Performance team, because of potential adverse impact to their end users, or to other customers. Before we formalized the load testing framework with SOASTA and Keynote, the load tests would require a good amount of preparation and each test would be subject to approval by the performance team. We ensure load testing providers are not firing all traffic from one or a limited number of locations to represent accurate traffic patterns from a wide variety of geographies. This ensure "real to life" caching and offload metrics, and performance measurement with Akamai's ability to route around internet latency.
SOASTA more recently has become prevalent in many of our larger retail customers. Last year we worked closely with them to make sure we trained their team and vice versa. They even made changes to their platform to ensure they could work well with us. The Akamai Performance team then signed off on a framework that is "pre-approved" for working with SOASTA and Akamai. The framework ensures full visibility with monitoring and reports on the Akamai side.
[Lelah] The click stream that is tested is essential for a valid load test. How do retailers identify the clickstreams through their site? How many are tested and how often?
[Paul] Retailers typically leverage analytics tools to identify common paths. They also identify high-risk problem areas based on data from previous performance monitoring tests. For example, features of checkout or checking out directly from a wish list. They then take these common and high-risk paths and then schedule them for load tests.
Pre-holiday they do testing with each release in lower environments. They start production testing in August and do it regularly up until holiday. Those who do it regularly and properly end up conducting between 8 and 20 tests.
The intention is to not take down the site. Scheduling the tests during consistently low traffic times is least likely to impact their end users. They include the real user traffic numbers in their load testing generator targets.
[Lelah] How do retailers set those load testing targets?
[Paul] Usually they take Thanksgiving night, Black Friday morning or Cyber Monday best case projections from the business. This year retailers were testing with the expectation of 40 - 60% traffic growth at peak. If they can exceed forecasted load then they stay up for the holidays. The hard part is sustaining traffic without performance degradation. The page performance targets under load are driven by their business team. They have alerts that fire if performance drops below a certain level.
[Lelah] What are the most common vulnerabilities found through load testing?
Normally its database and application server issues. We'll then work with them to see if we can provide offload for those vulnerable areas on Akamai. It also helps them determine how well their load-balancing infrastructure is working. We'll double check that their disaster recovery center (or other failover scenario) is operating appropriately. Often times a SPA [Shopper Prioritization Application - Akamai's traffic throttling solution] test is paired with a load test to evaluate when SPA should be used and how they can maintain optimal load and performance in their environment.
Retailers considering load tests through Akamai should make sure to contact their account rep to see if their load tests comfort with the pre-approved framework.
Reaching a unique cookie conversion rate of 5.1% at the 11PM hour
Morning show broadcasts reported on the hottest Cyber Monday deals at the 9AM EST hour along with many online publications doing the same yesterday morning. But most online consumers didn't need any encouragement to get shopping - most continued with their clicking exercises right through the morning commute. Early browsing behaviors mirrored those of Black Friday quite closely. The only clear difference is that consumers chose sleep over shopping as illustrated by the 25% decrease in browsing activities at 12AM. Additionally, average page views dropped from 14.1 to 11, a decrease of 22% early Monday morning (12AM -4AM).
While browsing activity is a perfect metric for sales predictions, assuming consumers are researching deals, the shift in purchasing behavior is what puts dollars in online retailer's pockets. Most retailers utilize Cyber Monday to bring their offline promotions online -a whopping 85% reported by NRF.
Looking at purchasing behavior in the ADS Data Platform, the shift from a predominately off-line shopping day, Black Friday, to one of the largest online shopping days of the year, Cyber Monday is clearly illustrated in the graph below.
Midnight purchasing behavior was 12.5% below Black Friday's post off-line shopping sprees. The shift in Cyber Monday's favor came at 6AM EST at a 32% increase in online purchases as we returned to our post-holiday weekend lives. Lelah also noted that the 6AM hour was the turning point for exponential growth in peak page views. A slowdown in that growth occurred at 8AM EST hour and resumed at a 32% growth at the 9AM hour. The short slow-down can be attributed to morning commutes and some tired pointer fingers. (Full disclosure: I purchase new pair of boots at 8:30AM, the morning commute certainly did not slow me down! )
That slight decrease in purchase activity was the exception to the norm yesterday. In fact, from 3PM to 11PM, online spending activities were consistently 50% higher than those observed on Black Friday. Purchasing activity peaked at 9pm, 104% higher than Black Friday at the same hour.
What time did you make your Cyber Monday purchase? We'd like to hear more please email email@example.com, tweet @lyss3b, or add your comments below.
9PM Purchasing Peak 104% Above Black Friday
Yesterday, I projected Cyber Monday's purchasing activity to be 150% higher than an average fall day, the verdict: 193% increase.
In addition, my projection was that Cyber Monday purchasing activity will trump Black Friday by 25%, the results are in and the Akamai ADS Data Platform saw a 49% increase in online purchases take place yesterday.
Stay tuned for a post on hourly trends seen on the ADS Data Platform throughout Cyber Monday.
Elyssa Duboys is a Senior Marketing Specialist at Akamai