The marketing spend on digital advertising is enormous, having eclipsed television ad dollars in the U.S. in 2017. Among the gold rings retailers crave: getting as many shoppers as possible to their web sites. But if you want strong traffic, be sure your site can handle it. Peak traffic periods can occur at any time, though many are predictable: Valentine’s Day, Mother’s Day, during special promotions and of course, the big holiday season: Christmastime, starting with Black Friday.
Catchpoint benchmarks the online performance (load time, availability) of hundreds of retailers. Looking back to last Christmas, most did a good job keeping their sites or apps up and running smoothly on Black Friday through Cyber Monday. But as traditional as the holidays are the inevitable retail site crashes and slowdowns. There were several on Black Friday 2018, with a few name brands going offline or showing “we’re having problems” slides.
The complexity of today’s Internet and the impossibility of perfection is a given, but still, most organizations know they can’t squander the peak traffic periods that are among their biggest sales days. Based on our tracking of retail sites, we‘ve identified the weakest links that can wreak havoc on site performance during peak traffic:
Third Parties: These are site elements hosted by outside companies and beyond your own firewall and direct control. They might include site search tools, marketing analytics, news feeds or others. Tracking these on an ongoing basis is important, because if one external player has a problem, it can slow down your entire site. So if you notice occasional third-party slowdowns on normal days, you need to keep a close eye on those services on the heaviest days, especially since they’re likely supporting numerous other sites also under heavy load.
Solution: incorporate only those third-party services you really need during peak periods. Have backups in place and make certain third-party code loads after the main site content, so a malfunction won’t prevent your more essential content from loading first, and therefore won’t be noticeable to site visitors.
“Heavy” Pages: A simple method to ensure faster load times is to make certain page weight (amount of data loaded into a visitor’s browser) isn’t too large. Many companies “lighten up” during peak traffic days, usually by deleting excess images or graphics. This can result in significant speed gains. So trim the fat when you know a heavy traffic day is ahead.
Micro Outages: A micro outage is when your site or app is down only in a specific geography, or for site visitors served by one ISP or some other finite category. These outages or slowdowns don’t make the news, but our tracking shows they are among the most frequent form of downtime.
Solution: don’t monitor page load times using national averages as your metric. Focus your attention on where your highest concentration of site visitors are located, and scan all relevant geographies.
Cloud-Only Monitoring: The cloud is one of IT’s greatest advances, but it’s not the optimal solution in all cases, especially for monitoring. Tracking the digital experience of your site visitors should involve cloud-located monitoring among your vantage points, but it’s a mistake to only use the cloud for this purpose.
Why? Cloud-only monitoring does not accurately assess the end user experience because it monitors from the perspective of the cloud only and therefore doesn’t track the entire application delivery chain — all the various waystations such as CDNs, ISPs, wireless networks, etc. — that stand between your server and the user’s device. We’ve seen many examples of web performance problems resulting from elements in the application delivery chain that would have gone undetected if cloud monitoring alone were used.
Solution: Make good use of the cloud but also monitor from backbone, wireless networks and other vantage points that more accurately mimic what site visitors are experiencing.
Infrastructure Scalability: Whether your infrastructure is your own server, a virtual machine or container — server overload can be a core reason for a web site or app to go offline or slow under peak load. While this sounds like an easy one to handle, it may not be that simple. Sometimes traffic peaks so sharply and quickly that a company’s scaling plan may not be able to keep up.
Solution: make sure your testing replicates extreme load spikes and the differing types of load (mobile, desktop) as much as possible. We also suggest an architecture that supports elastic horizontal scaling, which lets you easily add new capacity on demand. This is one area where the cloud shines and can help you handle loads that exceed your expectations.
Each season has its own heavy traffic days or hours. With these categories in mind, know that comprehensive monitoring from the end user perspective is how you’ll find the particular weak links inherent in your system. It’s the definitive way to gain visibility into all the performance-impacting elements in the delivery chain, including those beyond one’s own firewall. Keeping a closer eye on these during peak periods and having rapid response backup plans can help protect revenues, and enhance brand image and goodwill among site visitors, as increasingly more shopping moves online.
Mehdi Daoudi is CEO and Co-Founder of New York-based Catchpoint, a digital experience monitoring company. Prior to Catchpoint, he spent more than 10 years at Google and DoubleClick, where he made sure the infrastructure responsible for billions of transactions daily was always available and running fast. This inspired him to build a more comprehensive digital experience monitoring platform. He co-founded Catchpoint in 2008, which now serves hundreds of clients worldwide.