What One Second of Load Time Actually Costs an Ecommerce Store
Every speed pitch quotes a stat about seconds and conversion. Most of those stats are old, decontextualized, or from a study design that would not survive review. The cost is real, though. Here is how to think about it with numbers you can defend.
The canonical findings, with their caveats
The most cited public research comes from large retailers with traffic volumes most stores will never see. Their findings - that added latency measurably reduced conversions - were honest, but they were measured on their pages, their audiences, and their baselines. Treating a percentage from a 2009 experiment as a universal constant is where speed conversations go wrong.
What survives scrutiny across studies: slower pages convert worse, the relationship is steepest in the first few seconds, and mobile shoppers are the least patient. The exact slope is yours, not a benchmark's.
Compute your own cost per second
You need three numbers: sessions by load-time bucket, conversion rate per bucket, and average order value. Most analytics setups can produce this in an afternoon. Bucket sessions by Largest Contentful Paint - under 2 seconds, 2 to 3, 3 to 4, over 4 - and compare conversion rates across buckets. The gap between your fastest and slowest bucket, multiplied by the sessions you could move into the fast bucket, multiplied by order value, is your defensible dollar figure.
Two mistakes to avoid: attributing all of the gap to speed (slow sessions over-represent bad connections, cheap devices, and casual intent), and ignoring seasonality. A before-after read on the same traffic mix beats a cross-bucket read when you can get it.
Where the seconds hide
On a typical commerce stack, the biggest costs are rarely the platform itself. Third-party scripts - tags, pixels, personalization, chat - routinely account for the majority of main-thread time. Render-blocking fonts and unoptimized hero images follow. The audit order that pays: third-party JavaScript first, images second, fonts third, server response fourth.
The takeaway
One second is worth whatever your own bucket analysis says it is worth. For most mid-market stores the honest answer lands in the low single digits of conversion rate - which, against real revenue, is more than enough to fund the fix.