You are being asked to approve a platform decision that will shape the next five years of your business, and the pitch in the room has quietly changed. Two years ago the answer was composable. Break the monolith apart, wire together best-of-breed services, buy your way out of the constraints of a single platform. In 2026 the same vendors are more careful, because the bills from the first wave of composable projects have landed and they are not the numbers anyone promised.
Here is the shift worth understanding before you sign anything. Composable commerce is being repriced, not abandoned. The market has worked out that decoupling everything is a cost rather than a strategy, and that most brands are better served by decoupling almost nothing. The decision in front of you is no longer monolith versus composable. It is which parts of your stack genuinely earn their independence, and which are being sold to you because an architecture diagram is easier to present than restraint.
The first wave has reported back
For three years composable was sold as the direction of travel, the thing serious brands did once they outgrew an all-in-one platform. Enough of those programmes have now run their course that we can see the actual economics rather than the promise.
The mid-market replatform in 2026 lands somewhere between roughly £120,000 and £250,000 and takes five to ten months. Go composable or full headless and that becomes £400,000 to £1.5m and up, with enterprise migrations regularly reported in the £2m to £4m range. Those are not edge cases. They are the running rate for the architecture that was pitched as the obvious next step.
What makes the number move is rarely the part anyone budgeted for. The platform licence, the line everyone negotiates hardest, is only twenty to forty per cent of the total. The other sixty to eighty per cent is implementation, ERP integration, data migration, quality assurance at scale, and the orchestration layer that has to sit between all the services you have just chosen to separate. Composable does not remove that work. It multiplies it, because every service you decouple is a new seam that somebody has to build, test and own.
This is the honest version of the trade. The three-year total cost of ownership for a well-run headless build can come out twenty to thirty per cent lower than a traditional replatform, through cheaper infrastructure and faster release cycles. But the upfront cost is two to three times higher, and the saving only arrives if you have the engineering maturity to realise it. A brand that goes composable and then staffs it like a monolith gets the bill without the benefit.
The pitch was compelling for a reason. It promised freedom from a single vendor, the ability to swap any component for a better one, and a store that could move as fast as the team behind it. All of that is achievable. The quiet condition attached to it, the one that rarely made the slide, is that you become the systems integrator for your own business, permanently. That is a role, with a headcount and a salary bill, not a one-off project cost.
The licence was never the expensive part
The most useful thing a board can do this year is stop treating the platform choice as the decision and start treating the integration surface as the decision. That is where the money goes and where the risk lives.
When we replatformed Herman Miller's Colebrook Bosson Saunders business, the platform was the least interesting part of the conversation. The value was in getting the product data, the pricing logic and the back-office systems to behave as one coherent thing for the customer. That is the work that decides whether a premium brand feels premium at checkout, and it is the work composable makes harder before it makes it better.
The same pattern shows up on the operational side. On Swale Heating the wins came from a progressive rebuild that respected the systems already carrying the business, not from tearing the estate apart for the sake of a cleaner diagram. Every service you split off is a contract you now have to maintain between two vendors who did not design their products to work together. Middleware, orchestration and vendor management are the categories that most reliably surprise finance eighteen months in, and they are exactly the categories a composable architecture creates more of.
None of this makes composability wrong. It makes it a bill you should only sign when a specific part of your stack is genuinely holding back growth. The failure mode is buying the whole philosophy to solve one narrow constraint.
The platforms moved while everyone argued about architecture
The other reason the composable case is weaker than it was in 2024 is that the platforms did not stand still. The problem headless was meant to solve, escaping a rigid platform, has been partly solved inside the platforms themselves.
Adobe Commerce shipped 2.4.9 in May, with a modern PHP foundation and a rebuilt search layer, and it remains the right tool for brands whose complexity is real rather than aspirational. Shopify Plus has spent two years closing the gap, adding native B2B, a serious automation engine and its own headless stack, and now handles cases that would have needed a full custom build not long ago. The gap between "flexible but heavy" and "fast but constrained" is narrower than the composable pitch depends on.
For most established brands that means the interesting question is no longer whether to leave a monolith. It is whether the platform you already run, used properly, gets you most of the way there. A great deal of the flexibility brands pay composable prices for is sitting unused inside the licence they already hold. We spend a fair amount of time telling brands that the honest answer is to use what they have harder before they rebuild, and the systems and integration work that connects that platform to the rest of the business is usually the higher-return investment.
What a scaling brand should do now
Treat composability as a set of individual decisions, not a destination. Ask of each part of the stack a single question. Is this component holding back revenue or margin badly enough to justify making it a standalone system I have to integrate and maintain forever? For most brands the honest answer is yes for one or two components, usually the frontend or search, and no for everything else. That is selective decoupling, and it is where the real economics of the next few years sit.
Judge the decision on total cost of ownership across three years and the team you actually have, not the upfront licence and the team you wish you had. The 20-30 per cent saving is real, and it is conditional. If you cannot staff continuous delivery, the composable route costs more forever, not less.
Watch the direction of travel. The market has moved from "everything must be composable" to "most things should not be", and the vendors selling the architecture know it even when the pitch does not say so. The brands that win the next platform cycle are not the ones with the most modern diagram. They are the ones who spent on the seams that make a store feel like one business, and who resisted paying to separate the things that were working.
The platform decision feels technical, which is precisely why it gets made at the wrong altitude, by the people most excited about the architecture. It is a financial and strategic call about where your next few million pounds of build budget compounds. If you are weighing a replatform this year and want a clear-eyed read on what actually needs to change and what can stay, that is the conversation we are always happy to have.
