Greenfield can over-transform. Brownfield can under-transform.
What the ZEISS course correction really teaches about S/4HANA transformation.
Mission: BETTER! — What the ZEISS course correction really teaches about S/4HANA transformation
ZEISS did not suddenly discover that Brownfield is good.
Four years ago the company publicly explained why it wanted the opposite: clean up decades of grown R/3 complexity, harmonise processes across roughly 160 entities, stay as close as possible to SAP standard. That logic was compelling then. It may still be compelling now.
What changed, apparently, is the path.
And that is exactly the point. A target architecture is not the same thing as the next sensible step.
The record — and what is actually confirmed
This case is unusually instructive because we are not working from a single news alert. ZEISS documented its own starting position, ambition and method in public, over years.
2022 — the problem, in ZEISS’s own numbers. Around 14,000 SAP users. Roughly 53 million transactions per month. Four main SAP instances. Some 160 entities. And 78 variants of a purchase order. Around 300 business, process and IT experts working on the FIT4 programme. This was never a technical upgrade. It was a business transformation programme of the first order.
2022 — the decision. ZEISS decides “in principle” for Greenfield, because the grown R/3 structures were no longer fit for purpose, and because processes should be cleaned up, tool sprawl reduced, standards adopted. Notably, SAP’s own framing supports that reasoning: a conversion carries the existing structures forward, complexity included. The Greenfield logic was not naïve. It was reasoned. The move to cloud itself had been in planning since roughly 2020; Greenfield was the later, separate decision about method.
2024 — the method, on the record. CIO Carsten Trapp confirms the approach publicly: Greenfield, standard-first, a design authority, Microsoft Azure as the platform. Three pilots already rolled out, a fourth to follow — then the scaling plan: more than 150 entities by 2030, with business and technical readiness assessed before every roll-in.
Read that list again. Pilots. Governance. A design authority. Readiness gates. Executive sponsorship. ZEISS did most of what the textbooks — including mine — recommend.
May 2026 — the first cost signal. Carl Zeiss Meditec flags roughly €40 million per year to be offset through 2028/29 against rising infrastructure costs — driven by a new ERP and CRM system, the lease for the Jena high-tech site and higher shared-services charges within the ZEISS Group, counted together. SAP is not named in that disclosure, and no clean SAP number can be derived from the figure.
September 2026 — the course correction. Bloomberg, citing people familiar with the project, reports more than €200 million and a shift toward a conversion-based approach that preserves many existing data systems. ZEISS confirms it reconfigured the ERP migration “to achieve faster progress”. The investment figure ZEISS has not confirmed, and it declined to comment on the timeline.
Two clarifications before anyone runs with the headline.
First: this is not a retreat from cloud. ZEISS started moving SAP workloads to Azure in 2017; by 2024, about half of its SAP applications reportedly ran there, and ZEISS continues to use SAP HANA Cloud, Datasphere and BTP. “ZEISS abandoned SAP cloud” is simply wrong. “ZEISS changed the transformation method for its S/4HANA journey” is what the record supports.
Second: this is not a Public Cloud story. The programme runs S/4HANA on Azure. SAP itself is explicit that a system conversion into S/4HANA Cloud Public Edition is not possible — Public Edition requires a new implementation. Whatever ZEISS teaches, it teaches nothing about GROW.
The uncomfortable part: ZEISS had the facts
The cheap conclusion would be: with a better pre-study, this would not have happened.
The record does not support that. ZEISS had system data, documented process variants, a global template, four deliberately different pilots, a design authority, readiness assessments, change management, executive sponsorship. By any reasonable checklist, the fog had been cleared.
So the honest lesson is harder, and more useful:
The fog does not stay cleared
Transformation decisions should not rest on workshops, assumptions or architecture preference. Usage, cost, complexity and business relevance need to remain visible throughout the programme.
A decision that looked right on good evidence in 2022 can look different in 2026 — on new evidence: real delivery effort, real cost, real speed, real organisational load. Changing course on new facts is not failure. It is management judgment doing its job. What would be failure is refusing to re-decide because the first decision was announced.
That is the difference between a target and a next step. The target — standardisation, harmonisation, a clean core, less tool sprawl — may have been right all along. The step — rebuilding everything, at that breadth, at that speed, simultaneously — may have stopped being the right next move. Both can be true at once. From the outside, we cannot settle it. Which is precisely why the mechanism matters more than the verdict.
The warning nobody should skip
Here is where the market will get this wrong.
ZEISS justified Greenfield in 2022 with an argument that has not gone anywhere: the grown structures were the problem. A system conversion, as SAP itself describes it, is a largely as-is transition — existing processes carry over, custom code must be cleaned and adapted, data volumes should be reduced. And SAP’s own clean-core framing draws the line sharply: a new implementation starts clean; a conversion has to get clean.
That is an enormous difference, and it produces the sentence this whole case turns on:
Brownfield does not remove ECC thinking
A technical conversion can move the system to S/4HANA while preserving years of custom code, interfaces, process variants and organisational habits.
Just because something can be migrated does not mean it has earned the right to live on. Every object, process and interface in a conversion deserves an explicit disposition:
Retain it, because it is still commercially and technically sound.
Replace it, because today’s standard does the job as well or better.
Retire it, because nobody uses it — SAP now offers usage-based custom-code analysis for exactly this.
Decouple it, because it is genuine differentiation that belongs outside the ERP core.
A conversion without that disposition discipline is not transformation. It is inheritance.
Which yields the hardest sentence I can offer the post-ZEISS debate:
Overcorrection is not a decision
If an ambitious Greenfield programme proves too costly or complex, the answer should not automatically be an uncritical Brownfield. Both extremes can preserve the wrong things.
Three paths, not two — and a new gate
Three refinements follow for anyone using this case to sharpen their own decision model.
The binary is stale. SAP itself distinguishes three transformation paths — new implementation, system conversion, and selective data transition — and now provides a data-driven Transformation Approach Finder, built on usage and data profiling, to support exactly this choice. Selective data transition matters wherever the honest answer is “partly”: keep some processes, redesign others, take history selectively, consolidate systems. Where a conversion inherits too much and a new build changes too much at once, the middle path exists — and it is chosen on evidence, not preference.
The decision model needs an explicit gate between pilot and scale. Call it prove before scale. A pilot’s success is not the programme’s proof. A pilot’s job is to produce new facts about the programme:
actual implementation effort against expected
realised standard-fit against planned
true local deviations
true change load
updated cost-to-transform
updated time-to-value
an updated business case
And then a genuine management decision: is the path still right? Is the sequence still right?
ZEISS ran pilots, and by 2024 they were read as proof the template worked. But the step after the pilots was scaling to some 160 entities — a completely different risk profile. The pilot does not prove the programme. It produces new facts about it. The gate is where those facts get their vote.
The cost of producing facts is falling. Both refinements above assume the same thing: that establishing the facts — what the custom code actually does, what still carries business weight, what can go — is slow, expensive and bound to scarce experts. That assumption is moving. Semantic analysis of large, grown code bases is becoming machine-assisted, and vendors are already selling it as a category of its own. Discount the marketing, and one implication survives: when facts get cheaper to produce, a gate you could afford once becomes a gate you can afford repeatedly. Prove before scale stops being a checkpoint and becomes a cadence.
The economics of fact production change. The discipline of using facts does not.
What ZEISS should be remembered for
Not as the case that discredits Greenfield. Not as the case that vindicates Brownfield. The record supports neither, and both readings just swap one dogma for another.
ZEISS may have been right to choose Greenfield in 2022. It may be right to choose more conversion in 2026. The lesson is not which path won.
The lesson is that the path must remain challengeable.
A migration method is not a strategy. It is a consequence of a strategy, constrained by facts — and re-constrained every time the facts change. The facts choose the path. The project does not get to choose the facts.
That — not a preferred migration method — is what I mean by From Fog to Facts. And ZEISS adds the extension the model needed: not only facts before the factory. Facts before scale.
If every difficult Greenfield ends in Brownfield, we have learned nothing.
Your turn. If you were inside a programme that changed its transformation path mid-flight: what fact finally forced the decision — and how long had it been visible before it was allowed to count?
Sources & methodology. The original ZEISS transformation approach is evidenced by ZEISS’s own current SAP programme communication and a July 2024 CIO Germany interview with ZEISS Group CIO Carsten Trapp, in which he describes the Greenfield strategy, standardisation principles, pilot structure and global rollout approach. ZEISS corporate reporting provides the broader transformation context. The 2026 course correction is based on Bloomberg News reporting published 6 September 2026 and updated 7 September, including ZEISS’s confirmation that its ERP migration had been reconfigured to achieve faster progress.
ZEISS primary source — Greenfield strategy ZEISS selbst beschreibt die Strategie ausdrücklich als Greenfield und sagt, die SAP-Landschaft werde „from scratch“ neu definiert. Zeiss ZEISS – SAP: the interface between people and data
CIO Germany — Carsten Trapp on the S/4HANA programme Sehr starke Quelle für Greenfield, Standardisierung, Harmonisierung, Piloten, Design Authority und den geplanten Rollout auf mehr als 150 Entitäten. CIO DE CIO Germany – So migriert ZEISS von SAP R/3 auf S/4HANA
Microsoft / ZEISS — SAP on Azure Wichtig für die Abgrenzung: ZEISS hat SAP-Anwendungen bereits seit Jahren nach Azure verlagert; 2024 lief laut Microsoft etwa die Hälfte der SAP-Anwendungen in der Microsoft Cloud. Damit ist klar: Brownfield ≠ Cloud exit. Microsoft Microsoft – ZEISS migrates SAP applications to the Microsoft Cloud
Mission: BETTER! — independent executive perspectives on SAP Cloud ERP, Business AI and enterprise transformation. Grounded in experience, not marketing.
More of this work: Executive Clarity → linkedin.com/company/112524121
Independent publication. Not affiliated with or endorsed by SAP SE. SAP is a registered trademark of SAP SE.
First published on LinkedIn. Subscribe to Mission: BETTER! to get new editions as they appear — or browse the sapperment library.