Case 1: The company just below the threshold
Ramesh Engineering reports annual revenue of EUR 490 million. The bank's large-corporate threshold is EUR 500 million. If the model stops here, Ramesh remains in mid-market coverage. But the company operates through fifteen subsidiaries, uses a central treasury centre, runs a host-to-host payment factory and has a EUR 300 million syndicated facility.
The right decision is not to ignore the threshold. It is to recognise that the threshold is one signal inside a broader operating model. The relationship manager submits a governed override based on legal-entity complexity, financing scale, cross-border footprint and treasury sophistication. The override has an approver and review date.
The downstream consequences are clear. Global coordination is assigned. Dedicated implementation and service are available for the payment factory. Credit remains based on actual exposure and risk, not segment prestige. Existing pricing contracts are not rewritten automatically. This is what controlled judgement looks like.
Case 2: The large company with very simple needs
Sravanthi Foods has EUR 1.4 billion revenue but operates in one country. It uses three operating accounts, domestic supplier payments, payroll and one bilateral term loan. It qualifies as large corporate by size, yet its operational complexity is modest.
The bank can keep Sravanthi in large-corporate coverage because of financial scale and credit importance while delivering mostly standard products. A named relationship manager coordinates the relationship, but the bank does not force expensive dedicated technology or complex channels. Service can remain efficient with specialist escalation when necessary.
This case shows why segment and service tier can be separate. Segment determines strategic ownership. Service tier reflects operational need. Keeping these attributes distinct prevents over-servicing and makes cost-to-serve visible.
Case 3: The fintech that changes category without becoming huge
Malla Payments begins as a software company providing payment orchestration. It has EUR 60 million revenue and sits in commercial banking. It later receives regulatory authorisation to hold safeguarded client funds and initiates millions of payment transactions.
The important change is not revenue. Regulatory status, product needs and financial-crime profile have changed. The client should move to specialist financial-institution or regulated-fintech coverage according to the bank's taxonomy.
The new model brings different due diligence, safeguarding-account expertise, high-volume payments support, API specialists and institutional credit assessment. The re-segmentation is triggered by a licence event and recorded with effective date. This is a good example of an overriding classification rule.
Case 4: The project vehicle with almost no revenue
GreenRoad SPV Ltd is building a toll road. During construction it has almost no operating revenue but has EUR 700 million of project debt. A revenue-only segmentation engine classifies it as a small business.
The classification is economically meaningless. The SPV belongs in project-finance specialist coverage because lending structure, security, construction risk, concession agreements and controlled accounts define the relationship. The sponsor group can still be linked for commercial context.
The lesson for data architects is important: segment engines need entity-type and transaction overrides. Not every customer can be placed correctly using ordinary company-size fields.
Case 5: The subsidiary of a multinational
Guna Manufacturing opens a new subsidiary in Portugal. The entity has no historical revenue and five employees initially. The local bank system would normally place it in small-business coverage.
Commercially, however, the entity is part of Guna's global treasury relationship. It will use the group's payment factory, cash-management pricing and central service governance. The bank therefore inherits the global relationship segment for coverage while retaining entity-specific KYC, account and risk attributes.
The inheritance does not grant automatic digital authority. Users still need mandates. It does not automatically extend credit. It does not erase local product restrictions. This case demonstrates why relationship inheritance and control inheritance must be separated.
Case 6: The joint venture that should not inherit the parent model
Guna owns 50% of a battery joint venture with another industrial group. The bank serves Guna as a strategic multinational. An automated hierarchy rule assigns the JV the same strategic pricing and service rights.
That can be wrong. The JV is jointly controlled, has its own governance and may use banking relationships chosen by both shareholders. The bank should treat it as a distinct customer while recording the Guna relationship. Any parent pricing or access requires explicit commercial and contractual approval.
The segment engine therefore needs control-aware inheritance. A simple parent-child tree is insufficient.
Case 7: The public-sector utility
A city-owned electricity utility generates EUR 900 million revenue. It has commercial operations but government ownership, public procurement requirements and long-term infrastructure debt.
The bank could place it in large corporate based on size, yet public-sector specialists may add value because governance, procurement, guarantees and public accountability differ. The primary segment may therefore be public sector with an industry subtype for utilities.
Credit analysis should not assume sovereign support unless legally or economically justified. KYB should record state ownership and relevant control persons accurately. Segment guides expertise; it does not predetermine risk outcome.
Case 8: The insurer and the payment institution have the same revenue
An insurer and a payment institution each report EUR 400 million revenue. Both fall under the bank's financial-institution umbrella. Treating them identically would still be too broad.
The insurer may need custody, investments, liquidity and premium-collection services. The payment institution may need safeguarding accounts, settlement, APIs and large payment volumes. Their regulatory and financial-crime risks differ.
The model therefore uses a primary FI segment plus subtype. Coverage specialists, product propositions and due-diligence workflows can then differ while reporting still aggregates the broader institutional franchise.
Case 9: A client crosses the threshold because FX moved
A UK company reports GBP revenue. The bank converts it into EUR for segmentation. Currency movement pushes the translated amount from EUR 480 million to EUR 510 million even though the company's underlying business has not materially changed.
If the bank immediately transfers the relationship, the segment model becomes unstable. A better methodology defines a consistent conversion approach and transition band. The bank can require sustained scale or a material operating change before moving the client.
This illustrates why source data and calculation methodology are part of segmentation governance. A threshold is only as reliable as the data feeding it.
Case 10: Re-segmentation damages the client experience
A growing company is moved from commercial to large corporate. Internally, the revenue code changes overnight and the former relationship manager loses system ownership. The new manager has not yet been briefed. The client calls with an urgent payment issue and discovers that nobody is sure who owns the relationship.
The error is treating re-segmentation as data maintenance instead of migration. A proper handover includes open opportunities, service incidents, credit, KYC dates, key contacts and implementation work. The client is introduced to the new team before system routing changes.
Segment transitions should have effective dates and temporary overlap where necessary. Relationship continuity is part of successful classification.
Case 11: Service tier and segment point in different directions
A mid-market marketplace client sends 300,000 merchant payments each day. It has modest revenue but extremely high transaction dependency. The normal mid-market service model is not sufficient for incident coordination.
The bank keeps the primary segment based on commercial criteria but assigns an enhanced service tier based on volume and criticality. Dedicated operational monitoring and named service management become available without pretending the client is a multinational large corporate.
This is why independent attributes create a stronger model than one overloaded segment code.
Case 12: Segment politics threatens objectivity
A large-corporate team wants to retain a client that has sold most operations and now falls clearly within mid-market criteria. The team argues that the relationship is “strategic,” while the receiving team says the client should transfer.
Governance should use documented criteria, current complexity, expected growth and client interest. If an override is justified, it should be approved transparently. Revenue ownership incentives should not decide the classification.
Analytics can monitor excessive overrides by team. A high concentration of exceptions can indicate that incentives are distorting the model.
Case 13: Product code accidentally drives segmentation
A bank decides that every client using host-to-host connectivity is automatically large corporate. Years later, a standardised host-to-host service becomes available to mid-market clients, but the segmentation logic still treats the product as proof of large-corporate status.
The rule has confused product adoption with client segment. Segmentation inputs should represent durable client characteristics rather than temporary product architecture unless the product genuinely indicates complexity.
Rule reviews should therefore accompany product strategy changes. Otherwise old assumptions become invisible technical debt.
Case 14: Testing a segmentation release
The bank changes the large-corporate threshold and adds a regulated-fintech override. Before release, data teams run the new rules across the full portfolio and compare results with the current model. They identify how many clients move, which relationship managers are affected and whether any subsidiaries split unexpectedly from their groups.
Testers verify boundary values, missing data, effective dates and override expiry. Business teams review surprising moves. Only after impact is understood is the rule activated.
This is how a classification change should be treated: as an operating-model migration with measurable client and resource consequences.
Case 15: What world-class segmentation should achieve
The final objective is simple: clients with similar banking needs and complexity should receive an appropriate, sustainable service model. The bank should know why each client sits where it does. Relationship teams should be able to explain the classification. Systems should apply it consistently. Overrides should be visible. Changes should be controlled.
Segmentation fails when it becomes prestige, internal politics or a single turnover threshold. It succeeds when it helps the bank allocate expertise, standardise where possible, specialise where necessary and preserve a coherent client experience as companies grow and change.