There’s a wave of replatforming happening right now. Magento 1 reached end of life. Magento 2 has proven expensive to maintain. Salesforce Commerce Cloud is losing ground. And Shopify has made an aggressive push into enterprise, offering a compelling alternative for brands in the £20m to £150m range.
The business case often looks straightforward: lower total cost of ownership, faster time to market, a broader ecosystem of apps and talent. Boards are approving these projects because the numbers make sense on paper. The numbers on paper rarely survive contact with reality. Around half of major platform migrations go over budget or launch late, on most estimates. Some never launch at all. And when that happens, it’s rarely the agency who takes the blame.
I’ve been involved in replatforming projects from multiple angles: as the client leading the migration, as an advisor to brands mid-flight, and as the person called in when things have gone wrong. The patterns are remarkably consistent.
The three phases where projects fail
When I look at a replatforming project, I’m looking at risk across three phases.
Scoping
This is where the budget gets set on assumptions that won’t survive discovery. Requirements are gathered from stakeholders who don’t fully understand the current system. The agency estimates on what it’s told, not what it will find.
Build
This is where scope creep takes hold. Feature requests accumulate. The like-for-like mentality kicks in. Every stakeholder wants their thing included, and the agency says yes because every yes increases the project value.
Migration
This is where projects actually die. Data is messier than anyone expected. Integrations that “just worked” turn out to be undocumented custom builds. Go-live gets pushed while the team scrambles to fix what nobody planned for.
The accumulated yes
Development agencies aren’t to blame for scope creep. They’re responding rationally to the incentives in front of them. You ask for a feature, they quote for it. You ask for another, they quote again. Every yes increases the project value. Every no risks the relationship, or hands the work to a competitor who will say yes. The problem is that nobody in the room has the job of saying “not yet”, or “do we actually need this?” It’s the same principal-agent problem I see in agency retainers. The incentives are misaligned, and until someone names it, nothing changes.
The like-for-like trap
The most expensive sentence in eCommerce is “we want it to work exactly like the old site.” I understand the instinct. You’ve spent years refining your setup, your team knows how it works, your customers are used to it. But Shopify has opinions, strong ones, about checkout, product data structure and how apps integrate. You can bend it to your will with enough custom code and enough budget. Every customisation creates future fragility, and every workaround becomes something that can’t be updated without risk.
The migrations I’ve seen land well treat the replatform as a chance to simplify. What did we build five years ago that nobody uses? What custom feature has a native equivalent now? What can we stop doing? That requires someone willing to challenge the brief, and it won’t be the agency. They’re not paid to reduce scope.
Launch drag
The problem isn’t just the budget overrun. It’s the growth you’re not capturing while the project drags on. A three-month delay on a £40m brand isn’t just £100k in extra agency fees. It’s a quarter of trading on a platform you’ve already decided isn’t fit for purpose. It’s Black Friday on legacy infrastructure. It’s the roadmap items that can’t start until migration is complete.
I call the cumulative cost of every week beyond the original timeline launch drag, and for most brands I work with it runs between £50k and £150k a month. That number changes how you think about the cost of governance.
The missing role
Every replatforming project needs someone who understands both sides and has the board’s trust. Someone on the client side, not the agency side, whose job is to:
- Challenge feature requests against commercial priority. Not everything needs to be in v1.
- Say no, or not now, without damaging the relationship. The agency can’t do this.
- Translate technical trade-offs into business outcomes. Most boards don’t speak developer. Most developers don’t speak P&L.
- Protect the timeline when scope starts to creep. Someone has to hold the line.
Most brands don’t have this person internally. The Head of eCommerce is too close to the detail, the CTO too removed from the commercial reality, and a typical project manager doesn’t carry the domain expertise. So the role goes unfilled, and the project drifts for exactly as long as it stays that way. The technology was never the risk. The empty chair was.