Legacy-GTM Sprawl: Building One Compliance Data Layer Across SAP GTS, Oracle GTM, and OneSource, Without Ripping Them Out

What changed: GingerOS reads each GTM instance read-only where it offers an API and grants read access, and writes approved results back only to your ERP.

The short answer

Legacy-GTM sprawl is the condition where classification, duty, and origin data are trapped in separate, non-reconciling instances of SAP GTS, Oracle Global Trade Management, and Thomson Reuters OneSource that grew up around different acquisitions, regions, and ERPs. You fix it by laying one programmatic compliance data layer across those instances, treating each as a system of entry rather than the system of truth, so every instance reads one consistent classification and tariff record instead of holding its own divergent copy. GingerControl builds that layer as GingerOS, the AI trade compliance platform that runs on your process and the data you have, ERP or not: it connects to each legacy instance read-only and sits over the stack rather than under it.

Next step:Get the policy briefings

In this post
  1. Can you unify SAP GTS, Oracle GTM, and OneSource without migrating off them?
  2. TL;DR
  3. Why does legacy-GTM sprawl happen, and why is it so hard to unwind?
  4. How do you build one compliance data layer over the legacy stack?
  5. Why "system of entry, not system of truth" changes the reconciliation problem
  6. What the reconciliation actually looks like across three instances
  7. How does this compare to the rip-and-replace migration most vendors propose?
  8. Frequently asked questions
  9. Putting one data layer over your legacy GTM stack
  10. References

The guide to customs classification automation in SAP, Oracle and NetSuite shows how the external reasoning step fits alongside the ERP's item master.

Can you unify SAP GTS, Oracle GTM, and OneSource without migrating off them?

Yes. The augment-not-replace approach leaves the legacy GTM instances in place to keep doing what they already do, screening, document generation, filing handoff, and adds a thin data layer above them, built with GingerOS, that owns the authoritative HTS, origin, and tariff record that each system reads, while approved results are written back only to your ERP. No migration, no big-bang cutover, no re-validation of every downstream filing path.

TL;DR

Legacy-GTM sprawl is what happens when a multinational ends up running two or three legacy global-trade-management platforms at once, SAP GTS in one region, Oracle GTM from an acquisition, Thomson Reuters OneSource in another business unit, each with its own item master, its own classification of the same part, and its own duty logic, and none of them reconciling. The instinct is to consolidate onto one platform, but a rip-and-replace of a system that touches live customs filings is high-risk, multi-year, and the regulatory clock does not wait: SAP ended mainstream maintenance for SAP GTS 11.0 on December 31, 2025, and Oracle has announced that Premier Support for on-premises Transportation Management and Global Trade Management is ceasing, with the 6.5.x line the last on-premises release (SAP via Rimini Street; Oracle Support Doc 2966726.1). For a global trade compliance director governing 80,000 to 250,000 active parts across three legacy GTM instances and a dozen ERP plants, the augment-not-replace path is to build one compliance data layer over the stack rather than under it. GingerControl builds that layer as GingerOS, fitted to your process and connected read-only to each instance, with the Classification + Tariff Calculation API for systems that need a programmatic call, so the authoritative classification record lives in one place and every legacy instance reads from it, instead of each instance being re-platformed at once.

Why does legacy-GTM sprawl happen, and why is it so hard to unwind?

No one sets out to run three global-trade-management platforms. Sprawl accumulates. A company standardizes on SAP GTS in North America, acquires a business already live on Oracle GTM, inherits a OneSource deployment in a European entity, and a decade later the trade-compliance function is governing the same products across instances that were never designed to agree. Each instance was a reasonable decision in its moment. Together, they are the problem.

Before accepting inherited GTM records as a reliable classification baseline, use M&A customs due diligence to review the acquired company’s entry history and unresolved customs issues.

The reason it is hard to unwind is that a GTM instance is not a passive database. It is wired into live customs filing, denied-party screening, license determination, and document generation. Every one of those paths has been validated, audited, and connected to a broker or a filing channel. Ripping one out means re-validating all of it, which is why consolidation projects routinely run multi-year and stall when a reorganization or a budget cycle interrupts them.

The conventional answer, migrate everyone onto one platform, collides with three realities:

Reality Why it blocks rip-and-replace
Regulatory clock does not pause SAP GTS 11.0 left mainstream maintenance on December 31, 2025; after that date SAP no longer ships tax, legal, and regulatory updates for that release (Rimini Street summary of SAP timelines). A two-year migration cannot be the only answer to a compliance gap that is live now.
Filing paths are load-bearing Each instance feeds entry filing through a broker or channel that has been validated against your records. A cutover re-opens every one of those validations at once.
Vendor roadmaps are diverging Oracle is moving on-premises OTM/GTM customers toward cloud, with 6.5.x as the last on-premises release (Oracle Support Doc 2966726.1); Thomson Reuters launched ONESOURCE+ in November 2025 as an AI-powered network layer (Thomson Reuters press release, November 5, 2025). Betting the whole program on one vendor's migration path concentrates risk.

The augment-not-replace thesis starts from a different premise: you do not need every system to agree internally. You need every system to read from one authoritative record. That is a far smaller problem, and it does not require touching the filing paths at all.

Quotable insight: A legacy GTM instance is a system of entry, not a system of truth. SAP GTS, Oracle GTM, and OneSource were each built to do something with classification data, screen it, file it, document it, not to be the single authoritative source of it. Legacy-GTM sprawl is not three databases disagreeing; it is three systems of entry with no system of truth above them. Add the layer of truth and the disagreement stops mattering.

How do you build one compliance data layer over the legacy stack?

The data layer is not a fourth GTM platform. It is a thin, programmatic record that sits above SAP GTS, Oracle GTM, and OneSource, owns the authoritative HTS code, country of origin, customs value reference, and full tariff stack for each part, and gives every instance that record to read; GingerOS writes approved results back only to your ERP, never into the GTM instances. GingerControl does not sell a packaged "data layer" product; it builds this one as GingerOS, fitted to your process. The architecture has three moving pieces.

1. One authoritative classification and tariff record. GingerOS keeps each part's approved HTS code with the reasoning behind it: GRI logic, Section and Chapter Notes, and CROSS rulings. Where a system needs a programmatic call, the Classification + Tariff Calculation API takes a product description plus country of origin and returns a 10-digit HTS code with the U.S. duty layers in force (General, Special, Section 301, Section 232 with optional steel and aluminum pour country, and other Chapter 99 lines) in a single call. That record, not any individual GTM instance, becomes the system of record for classification decisions.

2. A reconciliation pass that reads what each instance currently holds, read-only, through each instance's API where it grants read access. Before the layer serves a single answer, it has to know where the instances disagree. The same part carrying three different HTS codes across SAP GTS, Oracle GTM, and OneSource is re-derived once through GRI reasoning, and the divergence is surfaced as an exception for a human to resolve, not silently overwritten.

3. A read-and-ERP-write-back contract. Each legacy instance reads the authoritative record through the export we set up for your project and keeps doing its job, screening, document generation, filing handoff, unchanged. GingerOS connections are set up per client: read-only, staged, with no master-data migration and no prebuilt connectors, and approved results are written back to your ERP, only where you approve, with an audit trail; GingerOS never writes into SAP GTS, Oracle GTM or OneSource.

An entity that runs on spreadsheets, with no ERP or GTM instance at all, still joins the layer: GingerOS runs on your process and the data you have, ERP or not, so we set up that entity's data storage and design its data flow. If its data is messy, we clean and organize it first, then build the AI system on it.

Here is the architectural contrast that defines the approach:

Approach Legacy GTM instances Source of classification truth Filing paths re-validated Time to first reconciled record Regulatory-clock risk Vendor lock-in Audit trail
GingerControl data layer (augment) Stay in place as systems of entry One authoritative record served by API No Weeks; go-live typically in 8 to 16 weeks Low; legacy stays compliant while the layer is built Low; layer is vendor-neutral over the stack Reasoning chain per record, one consistent basis
Rip-and-replace migration Decommissioned, replaced by one platform The new platform Yes, all at once Multi-year program High; gap persists until cutover High; concentrated on one platform Depends on the new platform
Status quo (stitched legacy GTM) Stay, each as its own system of truth None; each instance holds its own No Never reconciles Persistent; no single owner of accuracy Distributed across three vendors Fragmented across three exports

Bottom line: For a trade-compliance director governing 80,000-plus parts across SAP GTS, Oracle GTM, and OneSource at once, building one data layer over the stack delivers a reconciled, audit-ready classification record in weeks while the legacy instances keep filing, whereas a rip-and-replace migration leaves the disagreement live for the multi-year duration of the project. A single-platform migration is the right call only when the organization has already committed to one vendor's roadmap and can absorb a full re-validation of every filing path.

Why "system of entry, not system of truth" changes the reconciliation problem

The reframe matters because it changes what you are trying to fix. If you believe SAP GTS, Oracle GTM, and OneSource each ought to be correct on its own, then sprawl is an unsolvable governance war, three teams, three item masters, three classification histories, all defending their version. If you accept that none of them is the system of truth, the war ends. The instances do not have to agree with each other. They only have to agree with the layer.

This is also where reasonable care comes back into focus. Under 19 U.S.C. 1484, the importer of record is responsible for using reasonable care to enter, classify, and value imported merchandise. CBP's Informed Compliance Publication on reasonable care frames it as an operational standard, and 19 CFR Part 163 requires the supporting records to be kept for five years from the date of entry. When the same part is filed under one HTS code in the SAP GTS region and a different code in the OneSource region, an auditor does not have to prove which one is wrong, the inconsistency itself is the finding. A single authoritative record, with one documented reasoning chain behind it, is the cleanest demonstration of reasonable care a multi-instance shop can produce.

CBP's Reasonable Care guidance puts the obligation plainly:

"The importer of record is responsible for using reasonable care to enter, classify and determine the value of imported merchandise and to provide any other information necessary to enable U.S. Customs and Border Protection to properly assess duties, collect accurate statistics and determine whether any other applicable legal requirement is met." (CBP, Reasonable Care Informed Compliance Publication)

A data layer does not relieve the importer of that duty. It does something more useful: it makes the duty satisfiable at scale, because there is one record to stand behind instead of three to reconcile under audit. GingerControl is an HTS Classification Researcher; it follows the same reasoning process a licensed customs broker uses, GRI analysis, Section and Chapter Note review, and CROSS ruling research, and produces audit-ready documentation that supports the classification decision. The final 10-digit determination and entry filing remain the customs business of the importer or their licensed broker (CBP Ruling HQ H290535; CBP Ruling HQ H350722, January 16, 2026). The layer is the research foundation that every legacy instance reads from, not a replacement for broker judgment or for the GTM systems themselves.

What the reconciliation actually looks like across three instances

The most concrete way to see augment-not-replace is to follow one part. Suppose a wiring harness assembly is active in all three systems: HTS code A in SAP GTS (inherited from the original North American classification), HTS code B in Oracle GTM (set during the acquisition that brought the instance in), and HTS code C in OneSource (classified independently by the European entity). All three feed live filings. None of the teams knows the others disagree.

The data layer resolves this in a defined sequence:

  1. Pull current state. The reconciliation pass reads the part's classification, origin, and value from each instance and flags that three codes exist for one part.
  2. Re-derive once. The part is classified one time through GRI logic in GingerOS, producing a single candidate code with a full reasoning chain, not a vote among the three legacy answers.
  3. Surface the divergence as an exception. Where the re-derived code differs from any instance, a human reviewer sees the conflict, the reasoning, and the relevant CROSS rulings, and confirms the authoritative code. The system asks before it overwrites; it does not silently pick one.
  4. Make the one record available. SAP GTS, Oracle GTM, and OneSource each read the confirmed code through the export set up for your project, and GingerOS writes approved results back only to your ERP. Their screening, document, and filing functions run unchanged, now on a consistent input.
  5. Hold the line over time. Because the authoritative record is owned by the layer, the next time any instance drifts, the divergence reappears as an exception rather than disappearing into a silo.

For a team running this across an 80,000-to-250,000-part catalog, GingerOS classifies several products in parallel, at a volume agreed with each customer, and the API takes 120 items per call by default, sized to your throughput, so the initial reconciliation pass is a scheduled job, not a multi-year manual project. The point is not speed for its own sake; it is that the legacy stack never goes dark while the truth layer is built on top of it.

How does this compare to the rip-and-replace migration most vendors propose?

Most platform vendors answer legacy-GTM sprawl with their own migration: consolidate onto our cloud, decommission the rest. That is a legitimate strategy when an organization has already chosen a vendor and can fund a full program. But it answers a different question than the one a compliance director facing a live maintenance deadline is actually asking, which is "how do I get to one consistent, defensible classification record before my next audit, without taking my filing paths offline?"

The augment-not-replace layer and the migration are not mutually exclusive. The layer is the faster path to a reconciled record, and it is also the lower-risk bridge if you do eventually consolidate, because once one authoritative record exists, any future migration imports clean, single-sourced data instead of carrying three divergent histories into the new platform. You can build the layer now and decide on consolidation later, on your own clock rather than the vendor's.

GingerControl's role in this is deliberately narrow and honest: it supplies the classification record and the connections through GingerOS and the API, and leaves SAP GTS, Oracle GTM, and OneSource to do the operational work they already do well. GingerOS is fitted to your process, not a platform that demands you abandon the systems you have validated.

Frequently asked questions

What is legacy-GTM sprawl and why does it create audit risk?

Legacy-GTM sprawl is running multiple legacy global-trade-management platforms at once, typically SAP GTS, Oracle GTM, and Thomson Reuters OneSource, each with its own item master and its own classification of the same parts. It creates audit risk because the same product can be filed under different HTS codes in different regions, and under 19 U.S.C. 1484 that inconsistency is itself a reasonable-care finding. GingerControl addresses the root cause with GingerOS, which serves one authoritative classification record to every instance instead of reconciling three.

Can I unify SAP GTS, Oracle GTM, and OneSource without migrating off them?

Yes. The augment-not-replace approach builds a thin compliance data layer over the stack that owns the authoritative HTS, origin, and tariff record that each legacy system reads, leaving each legacy system in place to screen, document, and file as it already does. For a director governing 80,000-plus parts across three instances, this delivers a reconciled record in weeks rather than the multi-year timeline of a platform migration. GingerOS supplies that record and connects read-only to each GTM and ERP system that offers an API and grants read access, with connections set up per client and approved results written back only to your ERP.

Plan HTS classification API integration with ERP systems around input mapping, exception review and the approved value written back to the item master.

How does GingerControl decide the authoritative HTS code when three systems disagree?

GingerControl does not vote among the three legacy answers. GingerOS re-derives the code once through GRI logic, surfaces the divergence as an exception with the full reasoning chain and relevant CROSS rulings, and asks a human reviewer to confirm before overwriting. For a compliance team reconciling tens of thousands of conflicting parts, this is the difference between a defensible single record and an arbitrary tie-breaker; the system asks before it classifies rather than guessing.

Does building a data layer replace my customs broker or my GTM platforms?

No. GingerControl is an HTS Classification Researcher that produces audit-ready documentation supporting the classification decision; the final 10-digit determination and entry filing remain the customs business of the importer or their licensed broker, per CBP Rulings HQ H290535 and HQ H350722. The data layer also does not replace SAP GTS, Oracle GTM, or OneSource, they read one consistent record and keep doing their operational work on better input. GingerControl augments both broker judgment and the legacy stack rather than displacing either.

What happens to my records and reasonable-care position under this approach?

A single authoritative record with one documented reasoning chain is the cleanest reasonable-care demonstration a multi-instance shop can produce, because an auditor sees one consistent classification rather than three conflicting filings. 19 CFR Part 163 requires those supporting records to be kept five years from the date of entry, and GingerControl's audit-ready reports preserve the GRI citations, Section and Chapter Notes, and CROSS references behind each code. For a team facing a Focused Assessment across multiple entities, that consistency is far easier to defend.

How fast can a reconciliation pass run across a large multi-instance catalog?

An initial reconciliation across an 80,000-to-250,000-part catalog runs as a scheduled batch job rather than a multi-year manual effort: GingerOS classifies several products in parallel, at a volume agreed with each customer, and the Classification + Tariff Calculation API takes 120 items per call by default, sized to your throughput. The legacy GTM instances stay live and keep filing throughout, and GingerOS writes back to your ERP only the results your team approves.

Should I still consolidate onto one GTM platform eventually?

You can, and the data layer makes that decision lower-risk rather than foreclosing it. Once one authoritative record exists, any future migration imports clean, single-sourced data instead of carrying three divergent histories into a new platform, so you decide on consolidation on your own clock. GingerControl's role stays the same whether or not you consolidate: GingerOS supplies the classification record and the connections, vendor-neutral over whatever stack you run.

Some of our entities have no GTM system or ERP, only spreadsheets. Can they join the same data layer?

Yes. GingerOS runs on your process and the data you have, ERP or not: for an entity with no ERP, we set up the data storage and design its data flow, and messy data is cleaned and organized before the AI system is built. GingerOS reads emails, PDFs, Excel files, scans and photos, though not video, so those entities start from the files they already keep.

Putting one data layer over your legacy GTM stack

If your classification, duty, and origin data are trapped in separate SAP GTS, Oracle GTM, and OneSource instances that never reconcile, you do not have to choose between living with the sprawl and betting a multi-year migration against a live maintenance deadline. GingerOS lays one compliance data layer across the stack, treating each legacy instance as a system of entry that reads one authoritative classification record. Augment, don't replace.

A GingerOS project starts with a Prototype Week: one call, then a clickable prototype of your own flow in three days, with zero permissions and zero production data. See how a project runs →

References

  1. SAP GTS 11.0 End of Mainstream Maintenance (via Rimini Street)
    Data cited: SAP ended mainstream maintenance for SAP GTS 11.0 on December 31, 2025; tax, legal, and regulatory updates no longer provided after that date.
    Source: Rimini Street, SAP End-of-Mainstream-Maintenance Deadlines for GTS, BOBJ, BPC, BW and Hybris
    Published: 2025
  2. SAP GTS 11 End of Maintenance Migration Guidance
    Data cited: Post-deadline pathways (extended maintenance, customer-specific maintenance) and migration options including GTS embedded in S/4HANA.
    Source: SAP Licensing Experts, SAP GTS 11 End of Maintenance
    Published: 2025
  3. Oracle Support, Premier Support for On-Premises Oracle Transportation Management / Global Trade Management to Cease
    Data cited: Premier Support for on-premises OTM/GTM ceasing; 6.5.x line as the last on-premises release; Oracle transition toward cloud.
    Source: Oracle Support Document 2966726.1
    Published: Oracle Support knowledge base
  4. Thomson Reuters, ONESOURCE+ Launch
    Data cited: Thomson Reuters launched ONESOURCE+ on November 5, 2025, as an AI-powered intelligent compliance network spanning tax, trade, legal, and risk.
    Source: Thomson Reuters press release, ONESOURCE+ launch
    Published: November 5, 2025
  5. CBP, Reasonable Care Informed Compliance Publication
    Data cited: Importer-of-record duty under 19 U.S.C. 1484 to use reasonable care to enter, classify, and value merchandise; reasonable care as an operational standard.
    Source: U.S. Customs and Border Protection, Reasonable Care (Informed Compliance Publication)
    Published: September 2017 revision
  6. Recordkeeping requirements, 19 CFR Part 163
    Data cited: Records relating to an entry must be kept for five years from the date of entry.
    Source: eCFR, 19 CFR Part 163 (Recordkeeping)
    Published: current eCFR
  7. CBP Rulings HQ H290535 and HQ H350722
    Data cited: Classifying specific goods beyond the 6-digit HS level for importation constitutes customs business under 19 U.S.C. 1641 and requires a licensed customs broker; AI-assisted classification beyond six digits with Form 5106 (HQ H350722, January 16, 2026).
    Source: CBP CROSS rulings database
    Published: HQ H290535; HQ H350722 (January 16, 2026)
Your next step

Every system says something different about the same part?

Hand this to GingerOS.

Get the policy briefings Map one source of truth across your ERPs

Pick a 30-minute slot on the next page with Chen Cui, our CEO. Bring one notice or one SKU list (an ERP export, a spreadsheet or an email will do) and see what GingerOS does with it.

Reviewed by Michael Weick, LCB / CCS
Customs compliance manager with 42 years of experience (ex Subaru of America, Merck, and Motorola).

More in Global Trade

Our Risk-Free Promise

Full refund.

If you’re not satisfied before MVP acceptance, get a full refund of your deposit.

Click Me!

See if you qualify