Composable stopped being the decision some time ago. Gartner’s Magic Quadrant for Digital Experience Platforms, published in January 2025, carries a planning assumption worth reading twice: by 2026, at least 70% of organisations will be mandated to acquire composable DXP technology rather than monolithic suites, compared with 50% in 2023. Modular, API-first, independently deployable capability is no longer a differentiator in this category — it is increasingly what the category means.
So if you run a complex multi-site estate, the interesting question is no longer whether to go composable. It is who you assemble it with.
The word “partner” does double duty in that sentence — the platforms you licence, and the team who wires them together. In a composable world those two decisions have stopped being separable, which is why the seven factors below apply to both.
Every serious vendor now claims API-first modularity, so the claim itself tells you nothing. What separates one composable stack from another is the orchestration layer: how content, commerce, data and identity actually meet, who owns that code, and what happens when one component gets replaced.
Ask a platform where its boundaries are and what it expects to be handed. Ask a partner to sketch the integration layer of an estate they run, from memory. If they cannot, they have never had to maintain one.
The gap between “supports multiple sites” and “supports forty sites across fourteen markets with shared components and genuine local autonomy” is where most programmes come unstuck. What you want is a content model that lets a market inherit, override and diverge without forking the build. What estates often end up with is site one, cloned thirteen times, and thirteen quietly diverging codebases within two years.
Ask to see a live estate at your scale and complexity. Not a demo tenant.
Composable decouples delivery from content, which promotes the presentation layer to a platform choice in its own right — hosting, edge behaviour, build performance, preview workflow, and how quickly a marketer sees their own change. That is a real gain in speed and control, and a real addition to your supplier map.
The partner test here is simple: does the front end come from the same team that owns the experience platform, or have you just acquired two suppliers who will point at each other across an API boundary?
Gartner’s second planning assumption in that same Magic Quadrant is the one I would put on the wall: by 2027, 40% of organisations will fail to deliver impactful digital CX due to a lack of AI-driven intelligent content coordination and content operations strategy.
That is a content operations problem wearing a technology costume. Across a multi-site estate the failure is specific and recognisable — components used inconsistently, five versions of the same product description, brand drift by market, nobody quite accountable. Ask who owns the design system and brand governance the day after go-live, and what authority that person has.
Every DXP demo personalises beautifully. Eighteen months later, a lot of estates are running two rules, both written by an agency that has since left. Composable can make this harder before it makes it better, because the data layer is now yours to assemble rather than the vendor’s to bundle.
Ask what the smallest useful piece of website personalisation looks like on the proposed stack, who writes the next one, and what evidence exists that the last one worked.
Forrester’s Wave for Digital Experience Platforms in Q4 2025 was blunt about the direction: “Agents are now the center of the DXP. Not a feature. Not an add-on. The center.”
There are two implications and they pull in different directions. Internally, agentic workflows change how content operations work — which is only useful if your content is structured well enough for an agent to act on. Externally, the first thing reading your content is increasingly a model summarising your category for a buyer who may never see the page it came from. A fragmented estate that contradicts itself across markets hands an answer engine exactly the problem it hands a human buyer, and it gets resolved the same way: whoever explained it most clearly wins. That is why AEO and GEO have moved out of specialist SEO territory and into architecture conversations.
Licence cost is the number in the business case. It is rarely the number that hurts. The recurring spend in a composable estate is maintaining the connective tissue — version upgrades landing on five different vendor timetables, renewals falling in five different quarters, and the specialist knowledge each seam quietly requires.
Ask a platform what breaks when you remove it. Ask a partner what they have decommissioned, not only what they have built.
Seven factors, one underlying test. Composable architecture moves risk out of the platform and into the assembly. A monolith could be judged largely on its own merits because the vendor owned nearly everything, including the failure. A composable estate is judged on the joins — and joins are somebody’s ongoing responsibility, named or otherwise.
Which changes the shape of a shortlisting exercise. Rather than scoring vendors against a feature matrix everyone passes, work out which of these seven you cannot get a straight answer to, and treat that list as your real risk register.
We work across Sitecore, Optimizely, Contentful and Vercel, usually on estates carrying more sites, markets and stakeholders than anyone originally planned for. The pattern is consistent: the platform choice matters slightly less than most selection processes assume, and the operating model matters considerably more than almost any of them measure.
A composable digital experience platform is a set of independently deployable, API-first capabilities — content management, search, commerce, personalisation, customer data, front-end delivery — assembled into one experience layer rather than bought as a single suite. Gartner’s January 2025 DXP Magic Quadrant assumes that by 2026 at least 70% of organisations will be mandated to acquire composable DXP technology rather than monolithic suites, up from 50% in 2023 — so in practice most credible enterprise DXPs are now composable to some degree.
A monolithic suite gives you one vendor, one roadmap and one throat to choke, at the cost of flexibility in any individual capability. A composable DXP lets you select the strongest option for each capability and replace components independently, at the cost of owning the integration layer between them. The trade is not features for features — it is convenience for control.
Evidence at your scale, not at demo scale. Specifically: a live multi-site estate they run with shared components and market-level autonomy, a named owner for the design system after go-live, the ability to describe the integration layer without a slide, a track record of decommissioning systems as well as building them, and clarity on which supplier owns the front end.
No. Composability earns its keep when you have genuine variation to manage — multiple brands, markets, regulatory contexts or business models — and the internal capability to operate a multi-vendor estate. An organisation running three similar sites with a two-person digital team will usually get more value from a well-governed integrated platform than from a stack it cannot staff.
It moves responsibility. In a suite, the personalisation engine and the customer data usually arrive pre-connected. In a composable estate you assemble that data layer yourself, which gives you far better targeting ceilings and a much greater chance of shipping nothing at all. The determining factor is rarely the technology — it is whether someone owns the personalisation roadmap after launch.
Answer engines retrieve, parse and compress rather than browse, so a multi-site estate that describes the same capability differently in different markets is harder for a model to summarise accurately. Composable architecture helps if the content model is coherent and machine-readable across sites, and hurts if each site has been allowed to diverge. Forrester’s Q4 2025 Wave for Digital Experience Platforms placed agents at the centre of the category, which makes this an architectural question rather than a marketing one.
Integration maintenance, staggered vendor upgrade cycles, renewals spread across the year, specialist knowledge for each seam in the stack, and the governance overhead of keeping many sites consistent. Licence cost is the visible number; the connective tissue is where the recurring spend actually sits.
