German SAP architect Baris Özcan’s research finds post-go-live reporting failures are organisational, not technical. With fourteen months to the ECC deadline, name one owner per report chain before scaling offshore.

The technology research and advisory firm ISG published The State of SAP Migrations, drawn from a survey of 200 senior decision-makers at large global companies. Its breakdown is specific: 15% of the migrations studied finished on time and on budget, 27% missed either the budget or the schedule, and 58% missed both. ISG points away from the platform and toward how the programmes are run: multiple providers, unclear decision rights, loosely defined acceptance criteria, gaps in accountability, concluding that outcomes come down to governance, which enterprises underweight. Michael Dornan, ISG principal analyst, told The Register: “A lot of the delays are caused by people, not necessarily the technology.” Mainstream maintenance for ECC stops at the close of 2027, and around two-thirds of the SAP ERP users in ISG’s sample were still only planning, unlikely, The Register noted, to make the cutoff.

Those figures count the failures that show up on a budget line. A second kind does not: the migration lands on schedule, every indicator reports success, and the planners carry on working from the figures they trusted before cutover. That is the subject of a 2026 peer-reviewed article in the International Journal of Computer by Baris Özcan, a Dortmund-based senior SAP BW/BI consultant and architect who has worked both ends of the handover he describes: Organizational Models for SAP BW Project Teams in International Offshore and Onshore Delivery Structures. The standing assumption inside most programmes is that a report rejected after go-live was broken for a technical reason, somewhere in extraction or transformation logic. Özcan’s finding points to the delivery structure instead.

“The transports go through, the extractors behave, and the business is still working from its old figures,” Özcan says. “At that point you are not looking at a code defect. You are looking at a delivery structure where nobody was ever accountable for what the figure means.”

The industry surveys count how often this happens. Delays, testing gaps, and data problems recur in them year after year, and operations executives consistently name data quality as the thing holding back what they get from new systems. What the surveys do not describe is the route by which the defect arrives. Supply chain reporting is not one system but a sequence: extraction from source, transformation logic, reconciliation, comparison against historical figures the business already trusts, and sign-off by someone who plans against the result. On a large migration, each link belongs to a different team, often a different vendor, frequently a different continent. Every handover is a point where technical completion and business meaning can come apart, which is why Özcan argues team structure belongs in the explanatory model for reporting stability rather than in the project-management annex.

Özcan’s model turns on three continuities that a delivery structure either preserves or interrupts. Continuity of interpretation asks whether the engineer editing a mapping rule knows what that field means to the planner who reads the output. Continuity of validation asks whether testing reaches reporting semantics or stops once the technical result is correct. Continuity of ownership asks whether one accountable person holds a report chain end to end, from source document to the figure somebody acts on. Where all three are interrupted, the outcome matches what ISG observed from the outside: one team certifies that the transport succeeded, a second that the extractor behaves, a third that the infrastructure is ready, and the question of whether the business can rely on the output sits with no one in particular.

What the paper proposes in response requires no additional budget and depends on timing. It rests on two instruments, both of which must exist before an offshore or vendor team is scaled.

“I am not asking anyone for budget or tooling,” Özcan explains. “I am asking who owns this number, and asking it while the team is still small enough for the question to have an answer.”

The first instrument is an accountability register. Every business-critical report, inventory position, spend consolidation, supplier performance, and returns is listed by name, and each one is assigned to a single person accountable across the whole chain rather than to a team per layer. The second is a validation sequence that runs in a fixed order: extractor and transformation correctness first, source-to-report reconciliation second, comparison against trusted historical outputs third, validation during the stabilisation window fourth. A test plan that stops at the first stage has measured the system; the reporting is measured at the stages that follow.

Özcan has worked where these handovers multiply. At the Mercedes-Benz Global Logistics Center in Germersheim, he was the SAP business intelligence consultant on a service-parts programme running around 300 specialists, with development arriving from Argentina and Romania. The work covered catalytic converter recovery and after-sales returns reporting. Returns reached the system through web channels, third-party sales, company-owned workshops, authorised dealers, partner workshops and reverse logistics; the recovered metal was palladium, platinum and rhodium, so the figures carried real money. The architecture was tractable. The substantive work was semantic: a return meant six slightly different events depending on the channel it came through, and settling what the word meant, before anyone could trust what the report counted, was the actual deliverable. His work on it won him Innovator of the Year in the Digital Transformation category at the Cases & Faces 2026 International Business Award. The same shape recurred at BASF, where he led an offshore team of thirty on a consolidation bringing reporting from the Americas and Asia into one landscape, and at Volkswagen Financial Services, where he oversaw a landscape of some 2,000 process chains and 300 daily loads behind leasing, credit scoring and regulatory reporting, and coordinated the offshore team through the handover of live BI applications.

The questions are live in his current work. At Wieland Electric, where he works under contract, he has been validating SAP BW extractors across three ERP conversion waves ahead of an S/4HANA migration and assessing whether the existing BW data models are ready for it, the work that decides, before cutover, what has to be checked and in what order. An earlier assignment at Al-Khafji Joint Operations, the Saudi–Kuwaiti venture of Aramco Gulf Operations and Kuwait Gulf Oil, ran the other way: supply chain reporting that lived in Excel had to be rebuilt as a governed BW application, and as the business intelligence specialist of a three-person senior team he co-authored the blueprint that set the data model around source systems that had never been customised for it, working on site in Saudi Arabia.

The measurements change as well, and the ones he proposes are reportable on a dashboard: validation latency, meaning the interval between a technical fix and business acceptance; offshore tickets reopened after review; defects caught by source-to-report comparison before UAT closes; and credibility questions raised by users after go-live. Transport counts measure delivery rather than trust. KPMG’s 2026 supply chain research makes the same gap visible at industry scale: trust in the data and in the system’s recommendations is becoming the condition for autonomous operations, yet only 35% of organisations measure it. The firms that do, KPMG notes, treat trust as an operational capability, establishing clear governance, ensuring data quality and transparency, and creating accountability for AI-driven decisions.

For an organisation with a cutover booked before the deadline, four checks follow directly. Ask who on the programme can state, in business terms, what a disputed field represents to the person planning against it. Ask which layer the test plan reaches, and treat the stages beyond technical correctness as part of the plan. Ask for a named individual against each business-critical report rather than a team. And ask whether the non-productive environment has been accepted on all four of its dimensions or on privacy alone.

That last check is the subject of a further 2026 paper, published in Professional Bulletin: Information Technology and Security, Data Anonymization Approaches for Non-Productive Environments in Enterprise SAP Landscapes, and the sequence it describes is familiar to anyone who has sat through a stabilisation window. The regression suite passes on masked data in testing. Three weeks after cutover, a reversal posting arrives, the document chain it belongs to does not resolve, and the analyst tracing it finds the field that would identify the route replaced with a placeholder. Nothing was omitted. The rehearsal was run on data that could not stage that event.

Migrations cannot be rehearsed on production data carrying personal records, supplier terms and shipment routes, so teams mask, subset or synthesise, and a clean masking run becomes the readiness signal. Özcan’s argument is that a masking run answers one question of four. Substituted values can break joins, invalidate domain checks, and alter the document chains a regression test depends on, so the environment also has to be accepted as relationally faithful. The masked value is frequently the clue that would trace a discrepancy to a specific route, plant, or customer, so it has to be accepted as diagnosable. Transformed values alter cardinality and compression, so process chains run slowly enough to read as a regression or quickly enough to mask one, which makes performance comparability a separate acceptance. Each dimension carries its own owner; a privacy sign-off answers the privacy question. The states least likely to be exercised are reversals, delayed postings and unusual delta conditions, rare, and the conditions under which production incidents tend to surface.

Özcan has worked at both ends of that question, handling PII anonymisation in reporting queries at Messe Berlin and security-patch testing in the SAP BW landscape of the Federal Office of Bundeswehr Equipment, Information Technology and In-Service Support. Whether an existing landscape is fit to build on is also a question he has taken to practitioner audiences: he served on the expert board of the AITEX Summit Spring 2026, judging submissions across its four tracks, and sits on the council of the association behind it. The same reviewing work, assessing other practitioners’ architecture and code, underpins his fellowship of the Hackathon Raptors Association, to which five electors admitted him unanimously in 2026.

“Deciding who is accountable for the meaning of a number is an engineering decision as much as an administrative one,” Özcan says. “Settle it early and the migration arrives both technically finished and commercially usable. The 2027 deadline will test less of anybody’s SAP skills than whether an organisation can say, before cutover, who owns each number the supply chain runs on.”