with Miguel A. López-Morell (Murcia)

For anyone who has worked in retail banking, the history of Banco Nacional de México (Banamex) and particularly since 2014 is uncomfortably familiar. On going research with Miguel A. López-Morell (and do let us know if we can talk to you -in anonymity at bbatiz64 @ gmail com) tells that story in detail, and the lesson it draws is more precise than the usual narratives about legacy technology or multinational mismanagement.

A bank builds something that genuinely works — deep customer relationships, integrated systems, a culture of in-house expertise — and then watches it unravel not because the market changed, but because the organisation imposed on top of it was incompatible with what had been built. Our research aims to tell that story in detail, and the lesson it draws is more precise than the usual narratives about legacy technology or multinational mismanagement. Banamex’s decline was not caused by a single scandal, a single strategic error, or a single technology failure. It was the product of multiple pressures — post-crisis strategic reorientation at Citigroup, US regulatory liability exposure, and the Oceanografía fraud — all operating through a single critical mechanism: the imposition of an organisational architecture that was incompatible with the one that had made Banamex work.

The central argument

The paper distinguishes between three things that are routinely conflated in banking strategy discussions: core banking systems (the transactional engine), the platform (the operational layer connecting products and channels), and the architecture (the organisational logic that determines how information flows, who has access to what, and what the bank actually treats as its fundamental unit of value — the product or the customer).

The argument is that Banamex’s success as a universal bank between roughly 1970 and 2001 was not primarily the result of good technology choices or even good management, though both were present. It was the result of an architecture organised around a unified customer view. The bank built its systems — in-house, incrementally, over decades — around the principle that a customer was a single entity whose relationship with the bank spanned deposits, credit, insurance, and brokerage. That sounds obvious. In practice, almost no large bank actually builds its technology that way.

Moreover, as Carlos Marichal reminded us, thanks to the Legorreta family Banamex was not only a domestic technological trail blazer but the intermediary of choice for personal and corporate banking within Mexican elites, certainly those based in the capital and the Bajio region. Middle classes performatively followed suit.

What Banamex built — and where it came from

It is worth being precise about the origins of Banamex’s architecture, because they matter for understanding what was later lost. Banamex was not building something uniquely Mexican in isolation. Like most large Latin American banks from the 1960s onward, it was adapting broadly American computing approaches — transmitted through US vendor strategies, US-trained executives, and the general diffusion of computing practices across the Americas. What made the architecture locally distinctive was precisely that adaptation: the in-house modification of imported approaches to fit Mexican customer relationships, branch networks, regulatory conditions, and macroeconomic volatility. Scores of engineers were hired, trained internally, and assigned to systems for decades. When Citigroup later reimposed an undifferentiated American architecture, it was not introducing something entirely foreign to Banamex’s technological history. It was stripping out the institutional learning that had made borrowed approaches functional in a specific market — and that distinction is everything.

What Citigroup broke, and why

In the aftermath of the Tequila Crisis, Citigroup acquired Banamex in 2001. Citi inherited a platform more coherent than its own Mexican operation warranted. The initial integration reflected this: Citigroup’s smaller Mexican retail book was absorbed into Banamex’s systems, not the other way around. For the first several years, the acquisition worked reasonably well, precisely because Citigroup left the architecture largely intact.

The break came gradually and then suddenly, driven by forces that were simultaneously strategic, legal, and reputational. Post-2008, Citigroup reorganised globally around a product-led model — separate silos for cards, lending, deposits, and wealth management, each with its own systems and reporting lines. This is the dominant model in Anglo-Saxon banking, with real advantages for global compliance and capital management. What it destroys is the unified customer view.

The Oceanografía fraud in 2014 dramatically accelerated the imposition of this logic. But Oceanografía should be understood as a symptom and a trigger, not the underlying cause. The fraud exposed weaknesses in local controls, but Citigroup’s response was shaped as much by US legal liability as by operational logic. Under American law, senior Citigroup executives faced personal criminal exposure for misconduct occurring in foreign subsidiaries. The response — rigid centralised controls, dramatically reduced local risk appetite, decision-making transferred to New York — was overdetermined by that liability framework. As one senior executive recalled, after the scandal Citi “began to add absolutely rigid layers of control” that “eliminated flexibility and radically changed its risk appetite in Latin America.” Local executives lost agility. Decisions that previously could be made in Mexico City now required approval in New York. Products that could previously be tailored to Mexican customer segments now had to conform to global templates.

The result was what engineers inside the bank called spaghetti architecture: functional, barely, but impossible to modify quickly and ruinously expensive to maintain. One former executive described the cumulative effect with precision: “Every new management team arrives and starts introducing a new architecture, but nobody cleans up the previous one. They generate layer upon layer, and in the end they have absolutely brutal technological debt.”

The competitive consequence

Banamex had held roughly 22–23 percent of Mexican banking assets in the early 2000s, level with BBVA Bancomer. By the late 2010s it had fallen at least three positions in the national ranking. Its preeminance amongst elites, middle class, and recently bancarised largely gone. Our paper attributes this decline not primarily to underinvestment in absolute terms, but to loss of agility, customer focus, and local decision-making speed.

The BBVA Bancomer comparison functions as a controlled experiment. Two banks, same market, similar scale, both owned by foreign multinationals. BBVA granted its Mexican operation substantially greater technological and commercial autonomy. Bancomer adapted faster to Mexican market conditions, maintained a more coherent customer architecture, and steadily widened its competitive lead. As a former Mexican finance minister and later board member at a competitor observed, “the contrast with CitiBanamex is dramatic and can be explained, in part, by Citi’s decision to micromanage operations from New York, whereas Bancomer continued to enjoy significant technological autonomy from its Spanish parent.” The divergence in outcomes is not coincidental. It is the direct consequence of two different answers to the same organisational question poised by Gould & Campbell or Bartlett & Ghoshal: does local architecture serve the parent, or does the parent serve the local architecture? And to paraphrase the great Alfred D. Chandler Jr., Does IT follow strategy, or does it constrain action? … You tell us.

Percentage of individual assets relative to the total Mexican banking system for selected participants, 1991-2025

Source: Jiménez Bautista (2012), Comisión Nacional Bancaria y de Valores and authors. Data as of September, 2025.

Why this matters beyond banking

The conversation our paper enters is urgent within retail banking, where digital-native competitors — Nu, Revolut, and others — are winning in the most profitable customer segments precisely because they were built from scratch around unified customer architectures, with no legacy systems to maintain and no silo politics to navigate. Incumbents’ problem is not the technology budget. It is the architectural debt that cannot be resolved by adding new layers on top of old ones.

But the finding travels beyond banking. One reader commented privately: “[Banamex] tells the exact same story when Walmart aquired Aurrera” in Mexcan food retailing and other said “like the Spanish, we did things differently [than Citi] at MetLife” in Mexican insurance.

But our story travels further than Mexico. James Cortada, whose historical work documents IT-driven transformation across American industry for five decades and has been read by millions of people, identified the Banamex case as an instance of a general organisational phenomenon visible across sectors and countries. IBM encountered the same constraint internally during the 1980s and 1990s: accumulated IT architecture embedded strategic choices that could not be quickly overwritten by new leadership or new investment priorities. The mechanism is identical in retail, insurance, healthcare, and logistics. Any large organisation that has built substantial IT infrastructure over decades faces the same fundamental constraint: the architecture is not a neutral infrastructure sitting beneath strategy. It is the organisation. It encodes past decisions about what the institution values, what it can know, and how fast it can move. When new ownership or new management attempts to impose a different organisational logic without reckoning seriously with the architecture already in place, the costs accumulate invisibly for years before they become undeniable.

The Banamex workforce tells that story in a single statistic. In 1999 the bank employed approximately 38,000 people. By 2024, after two decades of Citigroup ownership, the figure stood at roughly 39,000 — surface stability concealing profound internal transformation. Following the decision to divest and the subsequent separation, internal estimates suggest the workforce will fall to approximately 31,000 by the end of 2026. That reduction represents the organisational cost of architectural misalignment made visible: two decades of layering incompatible logics onto a coherent foundation, and then the painful process of unravelling what was built.

Banamex was not destroyed by a fraud scandal, fintech disruption, or strategic retreat. It was undermined by an architectural mismatch that began the moment the logic of the acquirer was imposed on the logic of the acquired — and sustained by legal, regulatory, and competitive pressures that made correction progressively harder. That is a story worth understanding before it happens again, in banking or anywhere else.