10am, 4th of January, 2022. I’m on holiday with the family after an intensive 2021 growing StoreConnect during and after 2020. I was enjoying a moment of quiet when my phone exploded with alerts. One of our client servers was reporting a massive spike in traffic. Usually it processed an order every 15 minutes, it was now reporting an order every second!
There had been no email from our clients saying they were doing any major campaign and it was not a usual time for a massive spike in traffic… 4th of January isn’t a usual sale date. So I was instantly on high alert.
Moving away from the comfy chair on the balcony, I pulled out the trusty laptop and logged in to find one of our engineers already investigating. The traffic looked legitimate. One of our clients had put Rapid Antigen Tests up online for sale in Australia during a period when finding an Antigen Test to purchase in stock at a pharmacy was like winning the lottery.
Suffice to say the product was hot. Their SEO, StoreConnect autogenerated Google Merchant and Facebook Catalog feeds and marketing efforts worked superbly. They received a literal flood of very interested visitors.
What followed was 35,000 orders placed on their StoreConnect server in a little under 2 days. This gave an average of one order every four seconds over the two day period with many hours during the day time having many orders per second flowing through the store.
Their store went from an average of 3,000 orders per month to 35,000 orders in a couple of days. They also sold 20 times their usual monthly revenue in those same two days as well.
With no warning. No pre-warning of the caches. No pre-scaling of servers. Without even a second thought at scaling, they opened the doors and invited tens’s of thousands of customers to their store at once.
We host our systems on Heroku. Each of our medium and large clients have a dedicated Heroku app that scales independently of other clients. We built our system to automatically scale from the start. Spikes like these can come, unannounced in the eCommerce world.
Using the autoscaling facilities of Heroku our system added servers that handled the load rapidly. Then, after peak load had passed, they slowly scaled down. The site did not go down and, as far as we know, every client order was placed in a timely manner with no sale lost.
Our StoreConnect servers run independently of their Salesforce backend and synchronize constantly, which means that the added time spent by humans processing those 35,000 orders and the software triggers within Salesforce that handled allocation, stock etc, in now way slowed down the end user customer experience.
Our architecture proved incredibly valuable during this massive spike of orders, as some of the work flows this customer had put in place on the Salesforce side with the order object were taking up to 30 seconds to complete. It created a situation where the orders were being banked up in StoreConnect for over half an hour waiting for previous orders to complete saving before the new order could be moved over to Salesforce.
However, this backlog only affected the backend processing. The customer placing the order itself received instant service.
The delay in orders being processed within Salesforce itself didn’t really affect the client either, as their team spent a couple of weeks working through the backlog of shipping and delivering 36,000 orders. Additionally, as our platform architecture was designed for multiple channels of communication, when they ran out of the item, the platform marked it as unavailable and within seconds the site stopped selling the product.
So when clients ask us if StoreConnect can scale, I say yes, yes we can, and tell them about this event.
If you would like help to scale your online business, get control of your own customer data and move into the eCommerce 3.0 field, please get in touch.