Ask a senior engineer where their week went and you will rarely hear “building the thing we sell.” More often it is a patch that broke a backport, a workaround for a workaround, a fire-drill triggered by a dependency nobody dares to upgrade. Individually these look like ordinary maintenance. Added up across a quarter, they describe something closer to a slow leak in your company’s ability to move.

The invoice you never see

When leadership talks about the cost of a legacy commerce or subscription platform, the conversation usually starts and ends with the licence. That is the wrong place to look. According to replatforming analyses from BigCommerce and Elogic in 2026, the platform licence is only about 20-40% of a migration project’s total cost; the remaining 60-80% is implementation, ERP integration, data migration and QA. The sticker price is the small number.

The larger, quieter number is operational. The same body of industry analysis finds that legacy platforms silently consume 40-60% of every development sprint in workaround maintenance, backports and patch fire-drills. Read that as a capacity statement rather than a technical one. On a ten-person engineering team, something like four to six people are, in effect, permanently assigned to tending the past. You are paying senior salaries for senior people to keep yesterday’s decisions from collapsing.

None of this appears on a balance sheet as “technical debt.” It appears as slower releases, as a roadmap that keeps slipping a quarter to the right, as the feature a competitor shipped first. The cost is real; it is simply distributed across a hundred small delays instead of one visible line item.

Why debt compounds

The genuinely strategic problem is that this kind of debt behaves like debt. Each workaround becomes load-bearing. The next change has to route around it, which creates a new workaround, which the change after that must also respect. Over years the system accumulates a private logic that only a few long-tenured engineers fully hold in their heads — which quietly turns a technical risk into a people risk.

That is where the talent story enters. Strong engineers do not stay long in roles that are mostly archaeology. The work that retains them — designing, shipping, seeing customers use what they built — is precisely the work that a debt-heavy platform crowds out. So the platform that slows your releases is often the same platform slowly costing you the people who could speed them up. Speed-to-market and retention are not separate concerns here; they share a root cause.

A different default: toggle, don’t rebuild

The architectural counter-move is to stop treating capabilities as things you weld into the core and start treating them as things you switch on. VBWD, a full-stack self-hosted SDK, is built around an intentionally agnostic core with functionality delivered as plugins — payments, subscriptions, catalogue, CMS, booking, chat, and native MCP among them — each of which can be enabled or disabled without a restart. One Python backend drives a Vue/TypeScript web front end and native iOS and Android SDKs.

The point is not novelty for its own sake. It is that when a capability lives behind a clean seam, changing it does not require disturbing everything around it. There is far less to route around, and therefore far less of that 40-60% sprint tax to pay. Because the infrastructure does the heavy lifting, VBWD’s internal benchmark suggests a small studio can credibly run workloads normally reserved for large enterprises — a one-million-product catalogue of complex items importing in roughly 40 minutes, against 3-plus hours on a typical legacy platform. Treat that as a hedged, vendor-run benchmark rather than a guarantee; the shape of the claim matters more than the exact minutes. VBWD ships source-available under BSL 1.1, free for commercial use while annual attributable sales stay below the value of 6.7 BTC per year.

The honest caveat

Switching architectures does not abolish debt; it changes what kind you carry. A younger platform has its own rough edges. It has not spent twenty years absorbing every commerce edge case the way a long-established incumbent has, and some of those edge cases will matter to you. What you are really doing is trading a large, opaque, compounding debt for a smaller, more legible one — a debt you can read, reason about, and pay down deliberately rather than one that quietly grows while you tend it. For many organisations that is the better trade. For some, whose value lives precisely in those two decades of accumulated edge cases, it is not. The discipline is to know which you are before you move.

If you want the longer version of how this plays out in practice, our colleagues walk through it in how a small studio runs an enterprise commerce stack, and the architecture behind the toggle-not-rebuild model is documented at vbwd.cc/architecture.

If this maps to a conversation happening inside your own organisation, the useful next step is specific rather than generic: see your own workload on modern, self-hosted infrastructure. Request an enterprise installation and bring the numbers you’re trying to improve.

Sources: BigCommerce and Elogic 2026 replatforming cost analyses.