Black Friday is one of the most valuable times for retailers across the world because buyers shop until they drop. Marketing campaigns started already weeks ago and promised unbelievable discounts on all kind of goods. Those people who had something on their Christmas shopping list delayed their purchase for this special event. Shop until you drop is the slogan we hear so often during Black Friday sales.
This year, I’ve spent some time for collecting meaningful metrics during Black Friday sales of our leading Retailers websites in Switzerland. In this post, I will give you insights into the implementation, share some findings and potential loss of sales due to identified hotspots.
The Setup
I tried to keep it simple and focus on meaningful monitoring results. All retailers provide some kind of shopping process and comparing those would lead to huge deviations. Therefore, I’ve decided to just load their start pages and measure the time until all elements have been downloaded to my machine.
My monitoring setup consists of:
SaaS-based monitoring suite
Real browser-based simulation
Measured start page load time of leading Retailers in Switzerland
Executed measurement every 10 minutes
Collected response times, errors and replay log files
The Findings
My intention is to build awareness on the performance engineering & monitoring topic and not to blame hard-working retail companies. Therefore, I won’t tell you what organization has left behind and who is leading the field. Let’s focus on some problem spots.
# 1st Hotspot: Performance Spikes
The chart below shows that the issue has started already during the night on 23. November.
00:00 AM – first massive response time spike of 30 seconds
7:30 AM – response time jumped up to 25 seconds
1:00 PM – response time was still not acceptable
4:00 PM – response time went down to normal level
Obviously, 10:00 am to 4:00 pm is peak usage hour and retailers should try to provide responsive and reliable services during this extremely valuable period. For some reason, one retailer failed to handle the massive shopping activities and lost eventually a lot of sales revenue.
# 2nd Hotspot: Overload Situations
According to my monitoring data, some services were not able to deliver the expected results
Service reported overload situation
Detailed error information captured
The Impact
Shoppers expect response times of not more than 3 seconds. Research has shown that performance slowdowns leads to 10 % loss in sales because buyers are stopping to use slow websites.
The Swiss retailer above has lost too much money due to this slowdown during peak shopping hours. Hopefully, they will start their engines and prepare their websites for the next Holiday shopping season.
What to do next?
Performance is a journey and can’t be integrated overnight. It requires forward-thinking, experience, appropriate tools and a guiding hand which leads you through arising obstacles.
Some quick hints:
Holiday Checklist – read my next post on this topic coming soon
Load & Performance Testing
Monitor – Alert – Diagnose – Optimize
Be proactive and prepare your business applications month prior arrival of the next Holiday Shopping season.
Thanks also to ReactiveCharts for helping me with charting those Swiss retailer response time metrics.