Why the Data Layer Got Outsourced — and Stayed There
The two-decade equilibrium that nobody designed and nobody can unwind.
The first MRO master data remediation projects were sold around 2003.
Twenty-three years later, the same firms are selling the same projects to the same customers. The methodology has matured. The deliverables are more polished. The KPIs are better instrumented.
The drift back to baseline is the same.
In the first article in this series, I argued that SAP and IBM cannot solve the MRO data quality problem from inside their own platforms — that the schemas which won the enterprise have become impossible to refactor, and the bolt-on pattern is the only architecturally available response.
That diagnosis is structural. It is not the whole story.
The structural position of the platforms explains why they cannot fix the foundation. It does not explain why nobody else has, either.
For two decades, the work of fixing MRO master data has lived inside an arrangement that almost nobody designed deliberately, but that has held against demonstrably bad outcomes through six waves of platform releases, three generations of CIO turnover, and now a wave of AI investment that is running headlong into the same wall.
The arrangement is the SI services equilibrium. It is the second structural reason the MRO data problem has persisted.
This is not a story about bad SIs.
It is a story about an equilibrium none of the participants can exit alone.
How the Equilibrium Formed
In the 1990s, the platform vendors made a contractual choice that has shaped enterprise software ever since.
The vendor would provide the system. The customer would own the data.
That clause was reasonable at the time. ERP and EAM platforms were being sold as transactional infrastructure — they processed orders, scheduled work, tracked inventory. The data inside them was the customer’s operational record. It belonged to the customer, and the customer was responsible for its accuracy.
The clause has cultural weight even where it has lost technical sense.
When MRO data is bad — duplicate parts, missing classifications, inconsistent UOMs, broken vendor cross-references — nobody looks at SAP or IBM. They look at the customer’s data stewardship. They look at the SI’s migration job. They look at the consultant who set up the classification scheme.
The platform vendors are not blamed for data quality. They are not pressured to fix it.
Compare this to customer master data in CRM. Salesforce is blamed when duplicates proliferate. Salesforce has invested aggressively in match-merge primitives, deduplication features, and data-quality APIs because the market holds them accountable.
The MRO equivalent does not exist.
The accountability got assigned to the wrong layer of the stack two decades ago, and the assignment has held.
The Services Economy That Filled the Gap
When accountability sits with the customer, but the customer does not have the in-house capability to deliver, a market opens.
The system integrators stepped in. Accenture. Deloitte. Infosys. TCS. Wipro. Capgemini. And a long tail of MRO data specialists, each with their own methodology and their own cleansing factory.
The work was real. Customers needed it done. The SIs delivered it competently.
A typical MRO master data engagement looks like this. Six to twenty-four months. Tens to hundreds of consultants. Profiling, deduplication, classification, enrichment, governance setup. A defined scope, a defined deliverable, a defined exit.
When the engagement ends, the SI rotates off. The customer is left with a cleaner dataset and a governance framework on paper.
Eighteen to thirty-six months later, the data has drifted.
Not because the SI did poor work. Because the engagement was structured as a project, not as ongoing infrastructure. Projects end. Drift does not.
The customer hires the next engagement. Often from the same SI. The SI has institutional knowledge of the previous remediation, which is genuinely useful. The cycle resumes.
This is the engine. It has run for twenty years.
Why the Equilibrium Has Held
The most striking thing about the MRO data services equilibrium is not that it exists. It is that it has held against demonstrably bad outcomes for so long.
A customer pays eight figures every five to seven years for the same problem. The problem is, by every objective measure, not solved. And yet the same arrangement gets renewed, with the same firms, against the same scope.
How?
The equilibrium holds because every participant is locally rational. Each role inside it is doing the sensible thing given the constraints of that role.
The customer. In-house master data capability requires sustained investment in people and process that does not show up on a balance sheet as an asset. It is operational expense with no obvious ROI in any single quarter. Building it internally means hiring data stewards, training domain experts, instrumenting governance, and maintaining all of it through leadership transitions. Hiring an SI for a discrete project is far easier to justify, get approved, and explain to a board. The SI engagement converts a hard organizational problem into a manageable procurement problem.
The SI. MRO data work is profitable, repeatable, and politically safe. The same engagement methodology can be sold to dozens of customers in the same vertical. The work is sufficiently complex to defend against commoditization but sufficiently bounded to staff predictably. Recurring cycles mean recurring revenue. There is no commercial incentive to architect a permanent solution; the temporary solution generates better lifetime value.
The platform vendor. Partner ecosystem health drives platform stickiness. SI certifications, conferences, co-sell motions, reference architectures — all of it depends on the SIs having profitable practices around the platform. A genuinely-fixed-in-the-platform MRO data layer would cannibalize hundreds of millions in partner services revenue. The platform vendor is not consciously protecting that. But the gravity of the ecosystem is real, and it shapes what gets prioritized.
The CIO. The half-life of an enterprise CIO is roughly four to five years. The drift cycle for MRO data is five to seven. Most CIOs sign one remediation engagement during their tenure. They are not in the chair when the next one is signed. The recurrence is invisible at the level of the individual decision-maker.
Every participant is acting rationally inside their constraints. The equilibrium is the sum of those rational choices.
That is why it has held. And it is why none of the participants can unwind it from inside their own role.
Equilibria built from locally rational choices are stable. Stable does not mean correct. It means stuck.
What the Equilibrium Has Cost
A conservative estimate. Asset-intensive industries collectively spend $2 to $4 billion per year on MRO master data engagements globally. Across two decades, that is a forty to eighty billion dollar market.
The market exists because the work has to be redone, repeatedly, against the same data, by the same firms, for the same customers, with the same drift back to baseline.
This is not waste in the SI’s hands. They earned it. The work was performed.
It is waste in the customer’s hands. Every dollar spent on remediation cycle three is a dollar that did not buy the architectural fix that would have made remediation cycles four through ten unnecessary.
But the cost goes deeper than the dollars.
Every five-to-seven-year cycle resets institutional memory. The data stewards who learned the previous remediation rotate to other roles. The classification rules that were defined live in deliverables nobody opens again. The decisions that were made get re-litigated. The next cycle begins with a clean slate the customer paid to acquire and then paid to forget.
And now there is a new cost, one that did not exist for the first ten or fifteen years of this arrangement.
AI does not work on data that has been remediated three times and drifted three times. The agents are good. The orchestration is sound. The underlying records are still wrong, in the same ways, that the last three remediations were supposed to fix.
The CIO who signs the AI initiative is, in many cases, the same CIO who signed the most recent data remediation. The AI is supposed to deliver the value the data fix did not. It cannot, because the data fix did not stick, and the AI is now the third or fourth thing built on top of the same compromised foundation.
The equilibrium has produced a dataset that AI exposes more cruelly than any prior technology cycle. The data does not get worse under AI. AI just makes the cost of the data being wrong impossible to ignore.
The schema is the schema. AI does not change that. And the equilibrium that managed the consequences of the bad schema for two decades is now generating bills the AI initiatives cannot pay.
Why This Cannot Be Fixed by Better Execution
The most common response from inside the equilibrium is we will execute better next time.
Better methodology. Stronger governance frameworks. Tighter sponsor engagement. More rigorous KPIs. Continuous data stewardship after the engagement ends.
I have seen these proposals. I have seen them sold, executed, and renewed. They are sincere. The teams behind them are competent. The methodology improvements are real.
The drift comes back anyway.
Because the drift is not a methodology problem. It is a structural property of an arrangement where ongoing data quality is everyone’s responsibility and nobody’s job.
The customer does not own the layer; they own the data inside it. The SI does not own the layer; they own the engagement that produced the data. The platform vendor does not own the layer; they own the system the data sits in. There is no party in the arrangement whose ongoing accountability is the data quality itself.
When everyone is responsible for an outcome, no one is.
Methodology improvements operate inside this arrangement. They do not change it. They make each cycle slightly more efficient while leaving the cycle itself in place.
You cannot improve your way out of a structural problem.
What Has to Change
The equilibrium will not be unwound by any participant inside it acting alone.
The customer cannot unwind it without an in-house capability that is rarely justifiable in isolation. The SI cannot unwind it without abandoning a profitable practice line. The platform vendor cannot unwind it without restructuring its partner ecosystem and refactoring the schema, which the first article in this series argued they architecturally cannot.
What unwinds it is a layer outside the equilibrium that none of the participants currently provides.
A data layer that is owned, not engaged. Continuous, not project-based. Architectural, not methodological. Specifically built for prevention rather than remediation.
That layer cannot be sold as another SI engagement, because the engagement model is exactly what generates the drift. It cannot be built inside the platform vendor, because the platform vendor’s installed-base architecture and partner economics prevent it. It cannot be staffed by the customer alone, because the in-house investment required is structurally hard to justify in isolation.
It has to be a different shape of thing entirely.
A specialty infrastructure layer that lives outside the equilibrium and replaces the parts of it that have been quietly failing for two decades.
That layer is the subject of the third and final article in this series.
The pattern matters beyond MRO. The same equilibrium has formed around customer master data, financial reference data, and product information management in other categories. Each one has its own version of the locked-in services arrangement. Each one is now hitting the same AI wall.
MRO is the leading edge because the data is the most complex, the platforms are the oldest, and the asset-intensive industries cannot afford the cost of getting it wrong.
What gets built here will be built elsewhere.
The SI ecosystem will not disappear. It will reposition. The work that survives will be the work the in-house layer cannot do — strategic transformation, integration architecture, change management. The work that does not survive is the cyclical remediation factory, because the layer that replaces it makes the cycle unnecessary.
That repositioning is already starting. The SIs that recognize it early will be ahead of the firms that continue selling the old engagement against a problem the customer has stopped having.
The equilibrium that held for twenty years will not hold for another twenty. The question is which participants exit it deliberately, and which get exited.
This is the second article in a three-part series. The first article argued that SAP and IBM cannot solve the MRO data quality problem from inside their own platforms. This article argued that the SI services equilibrium that filled the gap has held for two decades through locally rational choices that are collectively stuck. The third and final article will describe the architecture of the layer that replaces the equilibrium.
About the Author
Raghu Vishwanath is Managing Partner at Bluemind Solutions. He has spent fifteen years building MRO master data infrastructure for asset-intensive industries.