In part 1 of our Black Friday blog post series, we investigated the importance of uptime – you don’t make any sales when your site is down, but also site performance – slow sites kill conversion. We gave a few hints about how e-commerce sites can improve availability and performance.
Now, with Black Friday 2017 behind us, let’s make a dive into the results and see how Swedish e-commerce fared. At Elastisys, we developed some tools that tracked the performance of Sweden’s largest e-commerce companies and some selected international e-tailers live during Black Friday. We studied uptime as well as site performance and found the following:
- 30 out of 230 major sites experienced as least one period of being overloaded to the point where they became unreachable.
- 127 out of 230 sites loaded at least 5 seconds slower than usual during some time of the day.
- On average, all 230 sites rendered 15% slower during Black Friday compared to a normal day.
In the below graph, we show in yellow the average site performance (the page load time in seconds) for all the e-commerce sites on a normal business day. In overall, site performance is good, with pages loading in 2s or less, with a slight slowdown during the evening hours when online shopping activity is higher than normal. In contrast, the green graph show the average performance of the same 230 sites during Black Friday – the most important day of the year for e-commerce. Many sites experienced significant performance problems straight off from midnight, with huge crowds of customers showing up only to be met by slow web sites and poor user experience. Similar problems occurred during the morning hours with a major peak at 10am, and a slowdown that continued all until midnight, when customer demand dropped off and site performance returned to normal.
The consequences: lost sales, lost customers, lost reputation?
In 2016, e-commerce revenue during Black Friday was 3.6 Billion SEK in Sweden alone. With a reported increase of 20%, this brings us to around 4.3 Billion SEK this year. We all know the frustration of slow web sites, and the strong correlation between e-commerce conversion and web performance is well studied. As a conservative estimate, Amazon found some years ago that each 100 ms delay resulted in 1% lost revenue. The average slowdown during Black Friday for the 230 sites we followed was 255 ms. Making the conservative assumption that revenue was evenly distributed throughout the day, this corresponds to 110 MSEK in lost sales due to bad performance for the Swedish e-commerce sector alone! Far worse than the single lost sale, poor site performance and user experience make customers reluctant to return, and can even result in bad reputation.
By studying how individual sites performed, we conclude that great care is needed when planning big campaigns. Quite a few e-commerce companies employed various types of Black Friday count-downs to midnight, just to experience instant overload when large crowds entered their site at the same time. Similarly, sending campaign emails or other messages to all customers simultaneously can create load spikes that are very hard to manage.
We also noted that a few e-tailers employed various types of queues, having customers wait to enter their website, to checkout goods, to pay, etc. Queues are good for entering a full flight, at a bus stop, and waiting for other scarce resources. In contrast, in e-commerce, competition is just one click away and customers are too impatient to wait.
Proactive resource management – how to recover those millions in lost sales
Elastisys provides a solution to these problems in the form of our predictive e-commerce management software. Elastisys’ software monitors the patterns of customers, uses predictive analytics to anticipate the future demand on websites, and automatically adjusts IT infrastructure to ensure an excellent shopping experience no matter what.
Elastisys software works by monitoring e-commerce platforms and extracting key performance metrics like number of visitors and infrastructure resource utilization. These metrics are combined with historic data and trend analysis to anticipate upcoming demand. Predicting future visitors allows Elastisys software to proactively take action to always keep your store available and delivering a great customer experience, even when marketing campaigns gets a lot of traction or when holiday offers or flash sales attract huge attention.