How Showpo Rebuilt Its API Layer for a Global Headless Migration

KrakenD gave us the scalability and reliability to grow our product offering with confidence, while cutting the operational overhead that was holding us back. – Showpo

The Problem: a monolith that couldn’t keep up

Showpo is a global fashion brand with a product catalog that moves fast. For five years, the platform ran on Salesforce Commerce Cloud, a setup that worked until it didn’t.

The core issue wasn’t any single integration. It was the architecture underneath them. Every third-party service, personalisation, reviews, content, returns, was wired directly into the frontend. There was no centralised control, no caching layer, and no isolation between external dependencies and the customer-facing experience. The dev team describes it plainly: “your typical monolithic codebase, slow and difficult to manage.”

For a brand that runs global promotional campaigns where traffic can spike sharply in minutes, that fragility had a direct business cost. Deploying a change meant touching the entire system. Adding an integration meant accepting a new point of failure. Scaling up meant exposing every backend to unfiltered traffic.

Something had to change.

The decision: going headless with a clean API layer

In 2025, Showpo partnered with Convert Digital to replatform to Shopify Plus with a headless Hydrogen storefront. The architecture would be composable: Shopify as the commerce engine, Nosto for personalisation, Yotpo for reviews, Sanity as the CMS, and a decoupled frontend that consumed all of it through a single, unified API.

That last requirement is where KrakenD came in.

A headless architecture only works if the API layer between the frontend and its backends is clean, reliable, and fast. Without that layer, a composable stack just becomes a different kind of spaghetti. Showpo deployed KrakenD as the Backend-for-Frontend (BFF), the single entry point that aggregates backend services, handles caching, and presents one coherent API surface to the Hydrogen frontend.

The deployment model was deliberately simple: a Docker image with the KrakenD configuration and license, synced directly to GCP. No separate cloud environment to manage, no infrastructure complexity to absorb.

The architecture after migration

Before · SFCC Monolith

Frontend

Commerce

Personalisation

Reviews

CMS

Each service wired into the frontend

After · Headless with KrakenD BFF

Hydrogen Frontend

BFF + Cache

Shopify

Nosto

Yotpo

Sanity

Single API surface, caching at the gateway

Where the old system had direct connections between the frontend and every external service, the new stack has a clean separation of concerns. KrakenD sits between the Hydrogen frontend and the backend services, aggregating four services behind a single endpoint and handling caching at the gateway level rather than pushing that responsibility into individual applications.

The frontend team no longer needs to understand what’s happening behind the gateway. The backend team can iterate without risking the user experience. And the operations team gets the visibility and control that the monolith never offered.

The results

The impact showed up across three dimensions: performance, deployment reliability, and day-to-day developer experience.

Backend load dropped by approximately 40%. By handling caching at the gateway layer, KrakenD absorbs a significant share of repeat requests before they reach internal services. For a brand that runs flash promotions and seasonal campaigns with unpredictable traffic spikes, that buffer has a direct effect on site stability.

Deployments are 20–25% faster and zero-downtime. Before the migration, each deployment carried real risk: occasional lag, manual environment setup, the possibility of a production incident. Now the process is: copy the config files and the license into the Docker image, sync to GCP. “It eliminates the need to manage separate, complex cloud environments,” one of the senior engineers noted. “It just works.”

Four backend services now resolve behind a single endpoint. The frontend consumes one API. What happens behind it is KrakenD’s problem, not the frontend team’s.

The engineering team’s assessment of the day-to-day experience is consistent:

The major benefit with KrakenD has been having a single, scalable gateway. It connects to all our backend microservices, handles caching, and significantly reduces the load on our internal systems.

It’s fast, secure, and the fact that it’s config-based makes it predictable. Having caching handled at the gateway level instead of at the application level has been a game-changer for our site speed.

What made the difference

Two things stand out from Showpo’s experience that are worth naming directly.

The first is the config-based model. KrakenD’s declarative configuration means deployments are predictable, version-controlled, and reproducible. Rolling back is a config change, not an incident response. For a team that previously operated in a system where changes carried high risk, that predictability matters.

The second is where caching lives. Moving cache responsibility to the gateway layer, rather than distributing it across individual applications, simplifies the entire backend architecture and gives the operations team a single place to reason about performance. For a fashion retailer where site speed during a campaign launch is directly tied to conversion, that is not a technical nicety. It is a commercial advantage.

Before and after at a glance

Showpo deployed KrakenD Enterprise as part of a Shopify Plus replatforming project delivered with Convert Digital. The platform serves customers across Australia, New Zealand, the United States, and the United Kingdom.