When an enterprise e-commerce platform breaks under load, the postmortem almost never blames the storefront. It blames the seam between the storefront and the OMS. Or the OMS and the ERP. Or the payment gateway and the order ledger.
The storefront is the smallest part
A storefront is a few hundred milliseconds of customer-visible latency on top of a long chain of inventory checks, pricing rules, promotion engines, fraud filters, payment authorizations, and order writes. The chain is the system; the storefront is the part that happens to be on screen.
This is why "replatform the storefront" projects so often disappoint. The pain customers feel — slow pages, broken stock, dropped orders — usually lives behind the seams, not in the storefront.
Three integration patterns we keep coming back to
- Outbox-driven sync between OMS and the storefront — never poll, never trust nightly batches.
- Idempotent payment-to-order writes with explicit reconciliation jobs running every few minutes.
- Marketplace channels as outbound projections of a single source of truth, not parallel realities to keep in sync.
Why this matters more during campaigns
Black Friday does not break new systems. It exposes the seams that were already brittle. Drop days, transfer windows, championship moments — same thing. Campaign engineering is mostly seam engineering, executed under time pressure.
“Storefronts get the credit when sales are good and the blame when sales are bad. The integration layer deserves both.”
What we ship instead of "another platform"
A unified integration runtime: outbox patterns by default, every channel addressed off the same canonical model, observability at the seam level rather than the page level. The storefront still matters — but it gets to be a thin, fast surface on top of a system that knows what is true.