Zero Downtime: Lessons from Retail Cloud Migrations That Worked

If you’ve ever been in a shop when the tills stop working, you know the feeling. That creeping awkwardness as customers shuffle in the queue, staff try to reboot systems, and someone starts calculating totals on a phone. Now imagine that happening online, at scale, during a seasonal sale. For retailers, downtime isn’t just inconvenient—it’s reputational damage, lost revenue, and frustrated customers, multiplied.

Even a short outage can hurt retail in tangible terms. In the Eurozone, IT downtime costs can reach around €4,600 per minute, and Global 2000 companies lose an average of €181 million per year to downtime, with each minute eroding profits and recovery times often stretching longer than in other regions.

In the push toward cloud modernisation, many retailers are promised the holy grail of "zero downtime." But what does that really mean—and is it even the right goal? At PCG, we’ve helped a number of retailers migrate their core systems to the cloud without disrupting business. The lessons we’ve learned are less about magical guarantees and more about smart architecture, strategic timing, and designing for resilience.

This article explores four cloud migration projects across the retail sector. Each one had different goals and constraints, but they all managed to avoid major disruption—and in doing so, offer practical insights into what works. As we walk through each, we’ll build a clearer view of how zero downtime isn't a fixed target but a shifting line—defined by context, capability, and customer expectations.

Redefining Zero Downtime

Before we dive into the real-world examples, let’s get clear on what “zero downtime” actually means. In the technical domain, it refers to systems staying continuously available with no interruptions but, from a business and end-user perspective, it’s more useful to think in terms of perceived downtime. Did customers notice? Did orders go through? Did any internal teams grind to a halt?

In reality, perfect uptime isn’t always necessary. What matters is ensuring business continuity during necessary changes. After all, the best migrations aren’t invisible because nothing got interrupted—they’re invisible because they didn’t break anything that mattered. This is the shift we’ve seen in the most successful retail migrations: from chasing perfection to designing for acceptable impact.

Let's be clear though: This is not a lazy approach that means settling for less. It means aiming for outcomes that are valuable and sustainable—using the right tools in the right places. Rather than stretching budgets and teams thin in pursuit of absolute perfection, it’s about applying effort where it delivers the greatest business impact. A well-executed, limited-scope change can often do more for continuity and performance than a sweeping rebuild with high risk and low return.

With that framing in place, let’s look at four examples that show how this principle plays out in practice—from infrastructure upgrades to platform re-architectures.

Four Retail Projects That Got It Right

1. Idealo: Integration Overhaul Without the Overload

Idealo, a leading price comparison platform, needed to streamline its complex partner integration process. Rather than a wholesale rewrite, the team focused on migrating specific components to AWS using a hybrid model. This allowed them to modernise step-by-step, prioritising improvements where the business would see the biggest gains.

The key was choosing not to migrate everything at once. By focusing on what actually needed to scale and integrating smarter with partners, Idealo improved data processing and reduced latency—all without disrupting the day-to-day platform operations.

Lesson: Don’t migrate for the sake of it. Migrate to meet specific goals.

This principle—purposeful, incremental migration—sets the tone for the examples that follow. Each case we’ll explore takes a similarly pragmatic approach, identifying critical functions and modernising them in isolation to reduce risk.

2. Bokbörsen: Better Email, Same Platform

Bokbörsen, Sweden’s largest second-hand book marketplace, relied heavily on transactional emails. Their legacy email system was fragile and hard to monitor but replacing it risked interfering with core services.

Rather than diving into a full platform overhaul, PCG implemented a new system using Amazon Simple Email Service (SES), Amazon Simple Notification Service (SNS), and AWS Lambda. This allowed Bokbörsen to log, monitor, and deliver emails reliably—without touching the main application. The result: better observability and fewer failed communications, implemented with zero disruption.

Lesson: Modernise the edges first. Low-risk upgrades can yield big wins.

Like Idealo, this project demonstrates that value often lies at the periphery. If core systems are stable but opaque or brittle, tackling supporting services—like communication or monitoring—can deliver meaningful improvements without triggering large-scale risk.

3. Fogia: A Smoother Order System, Piece by Piece

Fogia needed to replace a legacy ordering system that had become too rigid. But halting sales while swapping out the backend wasn’t an option. This challenge pushed the team to adopt an event-driven approach. Using AWS Lambda, Amazon Simple Queue Service (SQS), AWS Step Functions, and Amazon EventBridge, they created a modular backend that processed orders asynchronously.

“The use of AWS Step Functions, SQS, Lambda, and EventBridge allowed the new platform to decouple services, buffer data processing, and handle orders asynchronously—ensuring no disruption to users even as individual components evolved.”

This buffered architecture meant they could update or replace services incrementally without pausing the system.

Lesson: Use asynchronous architecture to isolate changes and avoid disruptions.

This builds on Bokbörsen’s lesson of isolating supporting services to reduce risk. Where that project decoupled email from the core system and achieved reliable communications without interfering with business operations, Fogia’s solution pushed this further—isolating internal logic using queues and workflows. It’s the same principle but scaled up: shield the business from change by building in buffers and hand-offs.

4. Re-board: Improving DCX with Resilient Architecture

In a project with Re-board, a sustainable retail display company, aimed at improving digital customer experience (DCX), the focus was on delivering real-time product feeds without risking instability. PCG used Lambda@Edge, pre-rendering, and robust fallback mechanisms to ensure the content was always available, even if backend systems lagged.

Instead of pretending failure wouldn’t happen, the system was designed to handle it. Even when backend components were temporarily unavailable, the system continued delivering a consistent and reliable experience to end users.

Lesson: Build for resilience, not perfection. Graceful failure is a success.

Here we see the final evolution of the same line of thinking: don’t just isolate failure—anticipate it. By designing for fault tolerance, the team made “zero downtime” about perception and experience, not system state. That’s where technical strategy aligns most cleanly with business reality.

5 Principles for Migration Without Mayhem

Looking across these projects, some clear patterns emerge:

• Start with the business priority. Idealo’s integration layer was the bottleneck; it became the focus.
• Pick the low-risk wins first. Bokbörsen modernised around the core, not through it.
• Decouple everything you can. Fogia made each system piece swappable.
• Design for failure. The DCX project didn’t deny risk—it absorbed it.
• Think incrementally. Every success here was achieved step by step, not with a big bang.

Together, these principles outline a working philosophy for real-world migration: not chasing perfection, but planning for reality—and ensuring continuity above all.

What 'Working' Really Means

Retailers don’t need fairy tales about zero downtime. They need strategies that work in the real world—during peak sales, unexpected traffic spikes, and Monday morning deployments. These migrations weren’t frictionless because nothing went wrong. They worked because the systems were designed with flexibility, visibility, and business continuity in mind.

That’s what "zero downtime" should really mean: Not "nothing ever goes wrong," but "nothing that matters breaks." From Idealo’s hybrid priorities to Bokbörsen’s peripheral upgrades, from Fogia’s event-driven buffers to the DCX team’s resilience planning—the pattern is clear: success lies in knowing what to protect, how to isolate change, and when to evolve.

If you're ready to modernise your retail systems, the first step isn’t a technical spec—it’s a business goal. And if you need help making sure it doesn’t break? You know where to find us.