Composable martech is an operating-model choice

Keeping a dependable marketing platform can be a mature architecture decision. Composable marketing technology, or martech, lets an organization connect and replace individual components, including a legacy application, without rebuilding its entire technology environment. A hybrid stack combines replaceable components with retained core tools. For a team that can maintain and govern that combination, a deliberate hybrid can be a lasting choice.

But that freedom brings continuing work because each connection creates maintenance, spending, and ownership decisions. The architecture must therefore fit the organization’s capacity to operate it while protecting the customer experience. Replacing every suite component offers little value when the resulting system demands more engineering, coordination, or oversight than the team can provide.

These operating choices shaped the November 2026 MarTech Conference panel, “Built to adapt: Why composable martech is becoming the new standard.” The discussion covered modernization at an achievable pace, deliberate governance, and trust across functions. Panelists approached those concerns through different questions: where data belongs, who owns the connected system, how to evaluate tools, and which responsibilities stay with the brand.

Decide which capabilities should remain stable

Modernizing at an achievable pace starts with deciding which capabilities should remain stable and which applications should be replaceable. Kristina Kalpokaite, head of martech at Loop Horizon, described a model that places customer data and core business logic in a central data warehouse, a shared system for storing and processing data. Specialized execution tools connect to that foundation to carry out marketing activities. Loop Horizon has a commercial interest in organizations seeking help with martech architecture and operations.

In Kalpokaite’s model, central management gives data quality, data hygiene, and audience segmentation a consistent home. Data hygiene keeps customer records clean and usable, while segmentation determines which customers belong in which audiences. With these functions managed centrally, teams can change an execution application without moving every related data responsibility at the same time.

Keeping essential functions stable also helps the business keep operating during modernization. Yefim Natis, principal analyst at Natis Labs, emphasized that business activity must continue while teams rebuild their technology. Today’s integration technology will eventually become legacy technology, he noted, so even a newly assembled architecture brings future maintenance and replacement work. Natis Labs benefits commercially from demand for technology analysis and advice.

How much replacement work a team can sustain also influences whether a hybrid should stay in place. Gene De Libero, founder of Digital Mindshare LLC, argued that retaining a hybrid stack can be a completed architectural decision. His business has a commercial interest in demand for martech strategy and advisory work. His recommendation makes the team’s ability to maintain and scale the system central to architectural maturity.

That capacity varies with staffing and regulatory demands. De Libero said lean teams may benefit from fewer integrations because each connection brings continuing maintenance obligations. Organizations with strict compliance requirements may likewise have good reasons to retain dependable core tools and limit change. For these organizations, a stable core is a deliberate response to operating constraints.

Okoone experts
LET'S TALK!

A project in mind?
Schedule a 30-minute meeting with us.

Senior experts helping you move faster across product, engineering, cloud & AI.

Please enter a valid business email address.

Assign responsibility for connections and spending

Once an organization chooses its architectural boundaries, it needs to assign responsibility for what crosses them. Separate applications exchange information with different expectations about fields, formats, and timing, while marketing activities depend on those exchanges working correctly. So flexibility at the component level requires a team that understands both the technical connection and the business activity it supports.

Kalpokaite emphasized that engineering and analytics partners need practical visibility into campaign execution because campaign intent determines which fields belong in which tools. An API, an interface through which applications exchange data, can work correctly while sending fields that fail to support the campaign. An engineer might, for example, transfer the requested customer attributes accurately even though they cannot support the intended workflow. Marketing must explain its purpose clearly enough for engineering and analytics to assess the request.

Understanding the campaign helps teams design useful connections, but the connected system still needs an owner. De Libero recommended treating the composable stack as an internal product and giving someone dedicated responsibility for it. That person must work across application boundaries because changing one component can affect data movement, dependent processes, and spending elsewhere. Otherwise, teams can manage their own applications while leaving the connections between them poorly governed.

System ownership becomes especially important when an organization leaves an all-in-one suite and takes on more integration work itself. De Libero identified data flows, dependency mapping, and cost governance as responsibilities the team must assume. Dependency mapping records which components and processes rely on others, so teams can assess what a proposed change will affect. Cost governance establishes who monitors spending and who can act when it changes.

Before mapping every possible connection, though, teams need to decide which information deserves to move. Craig Schinn, COO of Actable, advised identifying the customer signals that materially support engagement before building an integration. A customer signal is information about a customer that can inform a marketing action. Actable has a commercial interest in organizations investing in customer-data strategy and implementation, which is relevant to his advice about integration priorities.

Schinn’s approach gives each integration a business purpose and limits its scope. Transferring every available data point increases the information to manage without establishing that each item improves engagement. Selecting the inputs a workflow needs gives marketing, engineering, and analytics a shared basis for deciding what to connect. It also lets them judge whether the data reaching that workflow is dependable and useful.

Even a well-scoped integration can create a separate financial obligation when processing moves into a warehouse. De Libero warned that this change also moves compute consumption: queries, the requests that retrieve or process data, use infrastructure resources. Unmonitored queries can therefore increase infrastructure bills. Centralizing work requires a corresponding decision about who monitors those requests and the resources they consume.

Warehouse spending makes application choices a concern for finance as well as engineering. The stack owner needs to assess how a proposed workload affects infrastructure bills alongside application and integration costs, so a purchase may require both an integration assessment and a warehouse-cost assessment. Managing those costs as part of the internal product gives the organization a chance to act before a higher invoice becomes the first warning.

Evaluate tools through campaigns and production tests

These operating responsibilities should shape evaluation before a purchase. Platform renewals and vendor pitches can pull a team toward competing priorities, while a concrete campaign supplies a specific business requirement. Kalpokaite recommended starting with campaign use cases and identifying the exact data attributes each requires. Teams can then assess a proposed tool against the intended work and the data movement needed to support it.

The campaign-attribute test also exposes unnecessary synchronization. If an execution tool uses relatively few fields, copying a large database into it creates work and cost beyond the campaign’s needs. Kalpokaite’s recommendation helps the team judge whether that transfer is worthwhile. The ability to ingest large volumes of data becomes a capability to assess against a defined need.

A defined campaign requirement also gives marketing operations and finance a starting point for measurement. They should establish baseline revenue attribution models, agreed methods for assigning revenue contribution to marketing activities, as part of the initial architecture. Doing this early makes performance evidence available before renewal pressure arrives. Measured results and revenue attribution can then inform spending decisions alongside integration effort and operating costs.

But estimates of effort need to hold up in the working environment. De Libero encouraged teams to test vendor demonstration claims in their actual production pipelines, the sequences through which live data is processed and delivered. Testing should reveal the integration work and internal resources required before the organization commits its budget. A capability demonstrated by a vendor still needs to fit the buyer’s existing systems and staffing capacity.

Production testing is particularly useful when a team wants room to change an artificial intelligence (AI) capability later. Natis argued that composability lets an organization pilot a capability and replace it if costs rise or outcomes disappoint. The team can assess that application without making every other component part of the same decision. It must decide whether the ability to switch is worth the integration, staffing, and operating demands.

Keep control of customer data as AI tools change

The ability to switch applications also depends on what the organization can retain when a vendor relationship ends. Before adopting vendor-provided AI, teams should ask whether they will retain control of models and historical data, and whether important brand intelligence could become confined to a proprietary environment. Information and learning accumulate through use, which can make a seemingly replaceable application hard to leave. So evaluating an AI tool includes evaluating what can move with the organization.

That assessment must use the integrations available at the time of the decision. Kalpokaite stressed checking what already operates in production rather than treating a future roadmap promise as an available capability. A promised connection remains a future possibility until it works in the required environment. Building current operations around it would make the organization dependent on delivery it cannot yet use.

Yet available connections address only part of the control question because some responsibilities shape every application that uses customer information. Kalpokaite identified identity resolution, consent management, and customer data as foundations that should remain internally owned. Identity resolution determines which records refer to the same customer, while consent management governs permissions for using that customer’s data. Ownership of these functions influences how the organization recognizes people and acts on their information across execution tools.

Keeping those foundations internally owned gives the organization a consistent basis for evaluating new AI tools. A team can assess what happens to customer knowledge and essential brand intelligence over time while retaining authority over identity and permissions. The practical test is whether the organization can choose to change tools while keeping the information central to its customer relationships.

Give cross-functional teams authority to operate the stack

Retaining that control requires authority beyond any single application or department. Schinn argued that transformation needs alignment across marketing, IT, data, and analytics, with technology decisions coordinated toward sustainable business growth. The wider panel discussion likewise placed trust across functions at the center of modernization. Engineering’s responsibility for integrations and finance’s role in costs and attribution make their participation necessary to operate the chosen architecture.

Shared authority matters because the brand remains accountable as its suppliers and systems change. Natis emphasized that responsibility for customer experience stays with the brand even as tools, databases, and microservices, small, separately operated software services, change. A supplier can provide a capability, but the organization must retain the authority and capacity to meet its obligations to customers throughout those changes.

Key highlights

  • Match architecture to operating capacity: A hybrid martech stack can be a lasting choice when it fits staffing, maintenance, and compliance requirements. Marketing and IT should identify which core capabilities need stability before replacing applications.
  • Assign ownership across the stack: A dedicated stack owner needs responsibility for data flows, application dependencies, and operating costs. Finance and engineering need shared visibility into warehouse workloads because processing decisions affect infrastructure bills.
  • Evaluate tools against campaign requirements: Marketing operations should define the data attributes and revenue attribution measures each campaign needs. Engineering should test vendor capabilities in production pipelines to establish integration effort and resource requirements before budget approval.
  • Retain control of customer foundations: Keep identity resolution, consent management, and customer data internally owned. Before adopting vendor AI, confirm control over historical data, models, and accumulated brand intelligence, and verify that required integrations already work.
  • Give cross-functional owners decision authority: Marketing, IT, data, analytics, engineering, and finance need coordinated authority over architecture and operations. The brand remains accountable for customer experience as applications and suppliers change.

Alexander Procter

November 18, 2026

10 Min

Okoone experts
LET'S TALK!

A project in mind?
Schedule a 30-minute meeting with us.

Senior experts helping you move faster across product, engineering, cloud & AI.

Please enter a valid business email address.