What shared cloud reveals during big ecomm sales events

Cart abandonment starts climbing at 10:15.

The next few hours

You refresh the checkout analytics. The payment page is loading in 5.8 seconds, up from 2.1 two hours ago. You ping whoever handles the technical side, open a ticket with hosting support, and pause the second send batch while you wait.

The host responds at 11:20: server load is within normal parameters.

What follows is a few hours of everyone investigating the thing adjacent to the actual problem. The last plugin update gets rolled back to see if that’s it. The CDN configuration gets pulled and checked for caching issues. The payment processor gets called to ask if they’re seeing elevated volume from other merchants. By early afternoon checkout is back to 3.3 seconds. Not where it was, but the sale keeps running.

End of day

Traffic came in at 43% above your previous record. Conversion was down. Revenue landed at 69% of your projection.

Deloitte’s research on retail site performance puts a number on what that gap costs: a 0.1-second improvement in load time increases retail conversion rates by 8.4%. The difference between a 2.1-second payment page and a 5.8-second one isn’t a rounding error. It’s most of the distance between 69% of projection and hitting it.

The post-mortem

A week later produces a list: better pre-event load testing, a CDN audit, a conversation about checkout page optimization. Somewhere in a support ticket nobody goes back to read, there’s a note that server response time was elevated during the peak traffic window. That finding doesn’t make it onto anyone’s action items.

Nobody points at the hosting environment, because the symptoms don’t look like a hosting problem. They look like plugin issues, CDN capacity, payment gateway load. The hosting sits underneath all of those things, and most post-mortems don’t go looking there. Partly because the symptoms point elsewhere, and partly because there isn’t much to do about it when the environment is shared with however many other merchants had promotions running the same morning.

What the next event looks like

The next promotional event on your calendar has a higher traffic target than this one. The infrastructure underneath it is the same. The question worth building into the pre-mortem is whether the hosting environment is sized for the peak you’re planning, or the peak you hit last time.

About the author

[Name] is a [title] at Nexcess, a specialty cloud company that builds solutions for ecomm merchants whose peak days are supposed to be their best days. If you’re planning a promotional event and you’re not sure your current setup is sized for the traffic you’re expecting, that’s worth a conversation.