Opinion · MarTech economics · The discipline gapROI theatre
MarTech's case for spend has been written by marketing.
The CFO reading it as a value-engineering brief sees a different number.
Show a chief financial officer the renewal package for the marketing stack. They will see a number more than twice what was signed five years ago, attached to outcomes nobody can precisely pin down. They will ask a question the buying committee did not: what is this worth, against the case where the purchase had not happened? They will not get a clean answer. The case for spend has been written in marketing's language. The audit committee has been listening in another. The result is a category that has grown to enterprise-software scale while keeping the buying discipline of a tactical purchase.
Marketing has tried
Marketing has tried. It has been trying for thirty years. Return on Marketing Investment (ROMI) emerged in the 1990s as marketing sought finance-grade accountability. Return on Advertising Spend (ROAS) became the operational measure of the digital era. Marketing Mix Modelling (MMM) became the analytical school of the 2010s. Multi-Touch Attribution (MTA) emerged with the digital customer journey.
The strategy houses have built whole practices around the question. McKinsey runs a Marketing Return on Investment practice. BCG has its four-legged framework and its Smart Allocation tool. The Marketing Accountability Standards Board (MASB), created in 2007 to certify the rigour of marketing measurement, runs an audit protocol called MMAP (the Marketing Metric Audit Protocol). The frameworks exist. The literature is rich.
Why the frameworks do not survive audit
What is absent is the audit itself. On MASB's own record, no ROMI methodology in active commercial use has been independently audited under MMAP since 2007. Three decades of ROMI literature. Nearly twenty years of audit protocol. Zero certified methodologies. The frameworks have been validated inside marketing, against marketing's questions, on marketing's data.
This is not a failure of effort. It is a failure of design. ROMI, ROAS, MMM and MTA were built to answer marketing's question (which campaign worked, which channel pulled, what was the lift) rather than the CFO's question (what is the verified five-year value, against a counterfactual where the spend did not happen, after the full cost of ownership and the discount rate). Two adjacent questions, two adjacent disciplines. The gap is not in marketing's effort. It is in the wrong instrument.
McKinsey's 2024 CMO Survey makes the visible side of the gap quantifiable. Only 41 per cent of marketing leaders consider their own organisation mature in performance measurement. More than 70 per cent say they cannot adjust marketing spend dynamically based on effectiveness. Marketing has had three decades and a rich literature, and less than half of the function says it is there.
Marketing has been measuring marketing's question. The CFO has been asking a different one.
The cost the slide did not show
The financial reality is worse than the measurement gap. The number that gets approved is almost always the license. The five-year TCO tends to be a multiple of it. Implementation alone, for a meaningful enterprise rollout, runs between one hundred thousand and seven hundred and fifty thousand dollars before the system goes live. Integration costs compound with every additional tool, and the average enterprise now runs roughly ninety MarTech tools (Scott Brinker's chiefmartec landscape passed 14,000 in 2024). Professional services, training, ongoing operations, vendor management, failed projects and the opportunity cost of capacity tied up keeping the stack alive are rarely on the slide that approved the purchase. The license is the tip. The rest is the iceberg.
ROI theatre
The case studies that justify the spend are mostly produced by vendors. The lift figures cited are typically attribution-fragile and methodologically thin. Multi-Touch Attribution has been described, politely, as the platforms grading their own homework. Marketing Mix Modelling in its rigorous form is correlational rather than causal: it tells you spend correlates with sales, without proving the sales would not have happened anyway. Other capex categories of this size routinely apply discount rates, sensitivity analysis, counterfactual modelling and audit-grade methodology. MarTech has had almost none of it. The CFO who reads the ROI deck as a value-engineering brief sees a marketing document, not audit evidence.
The license is the tip. The integration, the services, the ops, the opportunity cost are the iceberg.
What value engineering would ask
None of this means MarTech investment should stop. It means the buying discipline should match the size of the investment. Value engineering, applied routinely in capex of comparable scale, asks five questions before approving spend. None are answered by ROMI, ROAS, MMM or MTA, because those frameworks were not built to answer them.
| The question | Why it matters |
|---|---|
| Verified value | What is the verified five-year value, against a counterfactual where the purchase did not happen? ROMI, ROAS and attribution measure differences inside the spend, not against its absence. |
| Full TCO | What is the full five-year cost including integration, services, change management and opportunity cost? The license is visible. The actual cost is a multiple. |
| Agent-era survivorship | What capability does this create that the agent layer will not subsume in eighteen months? Renewing the license without asking is paying twice for the same outcome. |
| Audit methodology | What methodology produced the ROI claim, and would it survive an audit committee? Most MarTech ROI claims are vendor-funded or attribution-fragile. |
| Failure cost | What is the cost of failure, and which line of the business carries it? Other capex categories assess this routinely. MarTech rarely does. |
There are architectures that survive these questions. They tend not to be relabelled CDPs or marketing-cloud bundles. They tend to be platforms priced against outcomes rather than seats, with lift measurement defined before go-live, an audit trail produced with every decision, a bounded path to first live use, and reference deployments whose percentages have been audited rather than asserted. The CFO who has read this page already knows which platforms in their evaluation set match this profile, and which do not.
When the agent layer takes the job
The third question on that list matters most for the years just ahead. It is also the question almost no procurement cycle is asking yet. The customer data architecture is in the middle of an inversion: agents traverse distributed sources and act where the signal lands. As that architecture matures, the functions some MarTech tools were sold to perform will be absorbed by the agent layer: segmentation, message composition, routing, small-scale decisioning. Some categories of MarTech will compress. Some will disappear. The CFO will be asked why the same license renews when the function has migrated. The answer "because marketing needs it" will not survive.
The buying decision that follows will not be marketing's alone. The CFO will be in the room from the start. Value engineering will return to the table. The architectures that pass the five questions have four properties the relabelling does not: an open architecture the operator can read and audit, reason-coded decision logs as a by-product of every decision, outcomes-aligned pricing against decisions made rather than seats licensed, and operator-controlled deployment so neither the data nor the audit trail crosses a vendor's boundary. Each property has been on enterprise wish-lists for a decade. In the agent era they become entry criteria.