Done Enough: The Case for Leaving a Process Alone

Most organisations are very good at identifying what should be improved. They are considerably less good at recognising when the process might benefit from being left alone long enough to discover whether the last five improvements actually worked.

In practice, many processes are not simply improved. They are redesigned, harmonised, remediated, standardised, transformed, and occasionally reorganised again before anyone has had sufficient time to understand what the previous version was trying to achieve.

The result is often familiar: more governance with less clarity, more controls with surprisingly similar findings, and teams quietly maintaining an ecosystem of spreadsheets, trackers, and local workarounds that nobody officially designed but everyone depends on.

In this article, I explore what happens when improvement starts interfering with itself, using examples from post-LIBOR finance transformation, GE Digital’s Predix programme, and post-2008 banking regulation, alongside research into organisational adaptation and change absorption.

I also propose a practical framework for a question organisations rarely ask explicitly: at what point is the highest-value intervention not another redesign, but simply allowing the process enough time to stabilise?

Curious whether others have seen a process cross the line from improvement into what might politely be described as operational turbulence.

When Improvement Doesn't Improve

There is a point where improvement stops being improvement, not because the process has reached some theoretical optimum, but because the organisation around it is no longer able to absorb the change without eroding the very outcomes the change was meant to strengthen.

This is not a philosophical concern. It is a pattern that has played out repeatedly in large, well-run organisations that were explicitly investing in better processes, stronger controls, and more disciplined governance.

Case Study: When Control Becomes Instability

In the years following the 2012 London Interbank Offered Rate (LIBOR) scandal, banks across Europe and the United States initiated extensive transformation programmes across Finance, particularly in Record to Report, with the explicit goal of increasing control, transparency, and auditability.

What made this case representative was not the scale of transformation, but the environment in which it took place.

Post-LIBOR, large financial institutions were operating under sustained regulatory scrutiny, where inaction carried visible risk, while action - particularly in the form of control enhancement and process redesign - was institutionally rewarded. In that context, each additional improvement was not only justified, but expected. The system did not lack discipline in identifying issues. It lacked a mechanism to declare stability as an outcome in itself.

One global bank, operating across more than 30 legal entities, became a useful illustration of what happens when improvement is not bounded.

Over a period of approximately four years, the R2R process was redesigned multiple times under different banners, including finance transformation, control remediation, ERP harmonisation, and target operating model alignment. Each intervention was rational in isolation, often triggered by a legitimate driver such as audit findings, regulatory expectations, system limitations, or post-merger standardisation efforts.

However, the cumulative effect of these interventions was not a progressively stronger process, but a continuously shifting one.

Process documentation was regularly updated but rarely fully absorbed by the teams expected to execute it. Control frameworks expanded in coverage but became harder to interpret in practice, as the relationship between controls and underlying activities was repeatedly redefined. Ownership was formally reassigned more than once, moving between global, regional, and local structures, which meant that accountability existed structurally but lacked continuity in execution.

At the operational level, teams adapted by creating informal stabilisers - shadow trackers, parallel reconciliations, and local conventions that allowed them to deliver outputs despite the instability of the formal design. These adaptations were not signs of resistance, but of necessity.

The paradox became visible in performance metrics. Control counts increased, yet audit findings did not decline at the expected rate. Standardisation improved on paper, yet cycle times remained inconsistent. Governance forums became more frequent, yet decision clarity weakened because the baseline kept changing.

The organisation had not failed to improve. It had failed to stop improving.

A Broader Pattern, Not an Isolated Case

General Electric and the Predix Platform

The same dynamic played out at a very different scale and in a very different industry during General Electric's push into industrial digitalisation, a case that has since become one of the most extensively documented examples of what continuous strategic redesign can do to an organisation that never allows its operating model to settle.

GE's digital ambitions began taking shape in 2013, when the company introduced Predix, a platform intended to become the operating system for the industrial internet - connecting machines, analysing sensor data, and enabling predictive maintenance across GE's vast portfolio of engines, turbines, and healthcare equipment. By 2014, GE was publicly reporting over one billion dollars in revenue from productivity solutions built around the Predix concept, and by 2015 it had formalised these ambitions by creating an entirely new business unit, GE Digital, with its own chief executive and a remit to make GE one of the ten largest software companies in the world by 2020. The target was $15 billion in digital revenue by that date, with Predix expected to contribute approximately half.

What followed was not a single failure but a cascade of strategic resets, each one a rational response to the problems created by the one before it.

The first structural problem was that GE built the organisation before it had built the product. As one Wall Street Journal post-mortem of the period observed, rather than assembling a small, focused team to develop a defensible product and then allowing the organisation to grow around it, GE created a large, expensive structure in anticipation of scale that had not yet been earned. GE Digital employed thousands of people and was expected to generate commercial returns from its first year, which meant that the pressure to bill other GE business units for internal work crowded out the longer-term investment required to make Predix genuinely useful to external industrial customers.

The second problem was that the platform's architecture was redesigned repeatedly as GE's strategic assumptions shifted. Predix was originally built on GE's own proprietary cloud infrastructure, including data centres that GE constructed and operated itself. This decision, made in order to control the end-to-end value chain, immediately put GE in competition with Amazon Web Services, Microsoft Azure, and Google Cloud - platforms that had years of development, far greater scale, and no dependency on GE's industrial relationships to attract customers. When this reality became apparent, GE migrated Predix to AWS, a transition that introduced significant delays, reliability issues, and customer confidence problems at precisely the moment when the platform needed to demonstrate commercial traction.

The third problem was that the platform itself was attempting to serve too many sectors simultaneously. Rather than establishing deep competence in one industrial vertical and expanding from that base, GE designed Predix as a horizontal platform for the entire industrial world - aviation, power generation, healthcare, transportation, and manufacturing - which dispersed engineering resources across too many integration challenges and made it difficult to develop the specialised functionality that any individual sector required. Lux Research, in a 2017 analyst assessment, concluded that Predix was not as fully developed as GE was representing it to be, had minimal market penetration, and had not been tested at the scale that GE's commercial claims implied.

Throughout all of this, the strategic framework within which GE Digital operated was in a state of near-constant revision. Under CEO Jeff Immelt, the ambition was to be a top-ten software company by 2020. When John Flannery replaced Immelt in 2017, he immediately called a two-month operational pause to assess what Predix had actually delivered, redefined the revenue target from $15 billion to $12 billion, and announced that future sales efforts would concentrate on sectors where GE had existing industrial relationships rather than attempting to win new industries. When Larry Culp took over in 2018, the strategic frame shifted again toward profitability, debt reduction, and a reorientation around GE's core industrial businesses. Each of these transitions was a legitimate managerial response to the circumstances that prompted it, but collectively they meant that the teams responsible for building and selling Predix were operating against a commercial mandate that had been materially redefined three times in less than five years.

The result was that GE Digital never operated a consistent strategy long enough to know whether any individual version of it would have worked. After approximately six years and an investment that by most accounts exceeded four billion dollars, and by some estimates approached seven billion when infrastructure costs were included, GE sold its digital asset business and scaled back to a narrower set of targeted, industry-specific digital applications - a model that commentators noted it probably should have adopted from the beginning.

The parallel to the R2R case is precise. In both instances, the problem was not that any single redesign was wrong. The problem was that the absence of a stabilisation period meant that the conditions required for one iteration to deliver value were dismantled before they had a chance to be tested.

The Regulatory Complexity Paradox

BIS, the Basel Committee, and the Post-2008 Landscape

A structurally similar dynamic is observable at the level of entire regulatory regimes, where the cumulative weight of well-intentioned improvement produces outcomes that run counter to the transparency and stability that each individual reform was designed to achieve.

Following the 2008 financial crisis, global regulatory bodies including the Basel Committee on Banking Supervision (BCBS) and the Financial Stability Board (FSB) undertook the most comprehensive overhaul of banking regulation in living memory. The Basel III framework, developed and phased in from 2010 onwards, introduced substantially higher minimum capital requirements, new liquidity standards in the form of the Liquidity Coverage Ratio (LCR) and the Net Stable Funding Ratio (NSFR), a leverage ratio backstop, and a set of surcharge requirements for globally systemically important banks (G-SIBs) that could, for the largest institutions, require total loss-absorbing capacity equivalent to 17.5 percent of risk-weighted assets. These changes were layered over the existing Basel II framework, which had itself added significant complexity to the predecessor Basel I standards, and were supplemented by national implementations such as the Dodd-Frank Act in the United States and the Capital Requirements Regulation and Directive in Europe, each of which introduced additional requirements, definitions, and reporting obligations that did not always map cleanly onto the international standards they were designed to implement.

The BCBS acknowledged this dynamic explicitly in a 2013 discussion paper on regulatory complexity, which noted that complexity within the regulatory framework is, to a degree, inevitable given that banking transactions involve nuances that resist simple rules, but that the pursuit of increased risk sensitivity had considerably increased the complexity of the capital adequacy framework, particularly in the calculation of risk-weighted assets - the central metric against which capital requirements are measured. The paper observed that while senior supervisors and senior bank management generally prefer frameworks simple enough to understand and apply with flexibility, the staff on both sides of the regulatory relationship who are responsible for ensuring day-to-day compliance tend to seek detailed rules to resolve the many ambiguous cases that arise in practice, and that this demand for precision generates ever-more-granular guidance that in turn creates new edge cases requiring further elaboration.

The consequence of this ratchet effect was that the Basel III framework, as fully implemented, was measurably more complex than the framework it replaced. Linguistic complexity - measured by document length and the density of defined terms - increased substantially between Basel II and Basel III. Computational complexity increased because capital calculations now required banks to run multiple parallel methodologies and use the more conservative output, a requirement that itself interacted differently with different bank business models in ways that were not always predictable. The BCBS evaluation of the reforms, published in 2019, noted that while the capital and liquidity frameworks both improved resilience, their interaction produced partially overlapping effects, and that the marginal benefit of increasing one standard diminished once the other was also binding - a result that was structurally invisible when each standard was designed in isolation.

At the level of individual banks, the operational implications were substantial. Compliance costs for large institutions increased sharply throughout the implementation period, with industry estimates suggesting that large banks were typically allocating between five and ten percent of their total operating expenses to compliance activities, with the largest institutions reporting annual compliance expenditures exceeding one billion dollars. Medium-sized banks faced proportionally similar costs. The European Systemic Risk Board, in a 2019 report on regulatory complexity, cited research estimating that Basel III implementation alone required a medium-sized bank to add between 135 and 210 full-time equivalent compliance staff. These were not costs that could be absorbed without consequence: resources committed to regulatory interpretation and reporting were resources unavailable for the operational improvement, customer service, and risk management activity that regulation was ultimately intended to support.

The deeper problem, though, was not cost but opacity. As requirements from different regulatory regimes accumulated - Basel III capital requirements interacting with Dodd-Frank stress testing requirements interacting with BCBS 239 risk data aggregation requirements interacting with the recovery and resolution planning obligations under the FSB's Key Attributes - the end-to-end logic connecting a bank's actual risk position to its regulatory capital calculation became increasingly difficult to trace. Where Basel I had used broad asset categories and fixed risk weights that any experienced analyst could evaluate quickly, Basel III allowed banks to use their own internal models to calculate risk-weighted assets, subject to regulatory floors and benchmarks that were themselves subject to revision. The result, as the BCBS discussion paper observed, was that capital ratios across banks became harder to compare, audit, and verify - not because banks were less transparent in their intentions, but because the framework itself had become too complex for outsiders to interpret with confidence.

This is the same phenomenon that appeared in the R2R case study, translated to a systemic level: each improvement made the overall picture less clear, not more. The pursuit of precision in risk measurement produced a framework that was less legible, not more, and the pursuit of comprehensive coverage produced an environment in which the relationship between individual requirements and the systemic stability they were collectively intended to produce had become genuinely difficult to evaluate. The FSB has in recent years acknowledged this, noting that as the focus of the international regulatory community shifts from standard-setting to implementation, the question of whether reforms have actually met their objectives - rather than merely expanded the volume of requirements- deserves sustained attention.

Where Continuous Improvement Breaks Down

The failure was not in any single redesign, but in the absence of a stabilisation boundary.

In each of these cases, the operating model assumed that improvements would accumulate. In reality, each intervention reset the conditions required for the previous one to deliver value. Without a defined period of stability, the organisation - whether a bank's finance function, an industrial technology business, or a global regulatory regime - never operated the same framework long enough to observe whether it actually worked.

Improvement was measured by implementation, not by sustained performance.

The Improvement Absorption Problem

Research in organisational adaptation, absorptive capacity, and continuous improvement sustainability has increasingly shown that organisations possess finite capacity to absorb change before responsiveness begins to undermine reliability. Sustainable adaptation depends not only on the quality of interventions, but on whether the organisation is given sufficient time to operationalise them before the next redesign begins.

In practice, improvement does not scale linearly. Early interventions often produce substantial gains by removing obvious inefficiencies, reducing variation, or strengthening control. Beyond a certain point, however, each additional intervention consumes increasing organisational energy simply to absorb the change itself. Teams spend more time relearning process logic, rebuilding coordination mechanisms, and recalibrating ownership structures than benefiting from the improvement being introduced.

At that stage, the organisation crosses from improvement into destabilisation. The limiting factor is no longer the quality of the redesign, but the organisation’s capacity to operationalise it before the next intervention begins.

What Continuous Improvement Disrupts

Every intervention in a process resets layers that are rarely accounted for in transformation plans - how individuals interpret the process, how exceptions are resolved in practice, how informal coordination fills the inevitable gaps in formal design, and how confident teams feel in their understanding of what constitutes a correct outcome.

These layers are not peripheral. They are the mechanism through which processes actually deliver results.

When changes are spaced too closely together, these mechanisms do not have time to stabilise. Knowledge remains partial, behaviours remain provisional, and ownership remains conditional on the next redesign. Under these conditions, improvement does not accumulate. It interferes with itself.

What transformation plans frequently underestimate is that every redesign resets organisational mechanisms that are difficult to document but critical to execution. These include tacit process knowledge, exception-handling conventions, informal coordination networks, ownership continuity, and accumulated confidence in how the process actually works under real operating conditions. None of these appear clearly in workflow diagrams or governance packs, yet they are often the reason stable processes continue to function despite formal imperfections.

The more frequently redesign occurs, the less time these mechanisms have to mature into reliable operational routines.

Why Organisations Struggle to Stop Improving

Most governance and transformation models are structurally oriented toward identifying and executing change. Improvement initiatives can be scoped, funded, tracked, and reported, which makes them legible to management systems. The decision to deliberately not change a process, by contrast, lacks the same visibility and is often interpreted as a lack of ambition rather than as evidence of sound judgment.

Very few organisations institutionalise the question of whether a process has reached a point where further improvement is no longer the highest-value action. The absence of this question creates a default bias toward continued intervention, even when the marginal benefit of change has already fallen below its operational cost - a threshold that, as the post-2008 regulatory experience makes clear, can be crossed gradually and without any single decision being obviously wrong.

The question is not simply whether a process performs well, but whether its performance is structurally repeatable without relying on unstable compensating behaviours. Viewed through that lens, organisations typically operate in one of four states.

The Stabilisation Test: A Practical Framework for Process Improvement

Before approving the next improvement, four questions are worth asking.

• Has this version of the process actually stabilised in practice - not just been documented and approved, but executed consistently without reliance on informal fixes that compensate for design gaps?

• Has the process completed at least one full performance cycle under real conditions, allowing its effectiveness to be observed rather than assumed?

• Are the issues being addressed true root causes, or are they symptoms introduced by the most recent round of changes, which would resolve themselves if the process were allowed to settle?

• What will this change reset that is not explicitly accounted for - including accumulated understanding, ownership continuity, exception-handling conventions, and trust in the process itself?

  
  

If these questions cannot be answered clearly, the highest-value intervention is not another improvement. It is time.

The Stabilisation Gate

One practical implication follows from this framework: organisations should govern process change in the same way they govern investment risk, architectural complexity, or control exposure. Not every identified improvement should proceed immediately simply because it is technically valid.

Before approving another redesign, leadership should be able to demonstrate four conditions:

• the current process has stabilised operationally;

• baseline performance is measurable and understood;

• the proposed intervention addresses a true root cause rather than disruption created by prior changes;

• and the organisation has sufficient capacity to absorb another transition without degrading execution reliability.

If these conditions cannot be demonstrated clearly, the highest-value intervention may not be another redesign, but a protected period of operational continuity.

Where the Discipline Changes

In the early stages of a process lifecycle, the primary capability is the ability to identify gaps and mobilise change. At a later stage, the more critical capability becomes the ability to protect stability long enough for the process to become reliable, predictable, and internally coherent.

This shift requires moving away from equating activity with progress and towards recognising that performance often depends on continuity rather than iteration. A simple principle captures this transition: no process should be changed more frequently than it can stabilise.

From Better to Enough

The bank in the earlier example did not arrive at this realisation through design, but through constraint. Change fatigue, operational risk, and diminishing returns eventually forced a pause in the improvement cycle. Only then did the process begin to exhibit the characteristics it had been designed for - consistency in execution, clarity in ownership, and predictability in outcomes. The process had not become perfect. It had become stable enough to work.

GE's eventual recalibration followed a similar path: only after the scale of the Predix ambition was substantially reduced did the remaining digital initiatives begin to show the kind of focused, measurable progress that the broader programme had promised from the beginning. The regulatory landscape is still working through its own version of this transition, as the Basel Committee and its member institutions begin to grapple seriously with the question of whether the complexity accumulated over a decade of reform is beginning to subtract from the transparency and resilience it was designed to create.

There is a difference between a process that is continuously improved and a process that is consistently performed. One signals effort and the other delivers results.

Most organisations are well equipped to pursue the first. Far fewer are structured to recognise and protect the second. The organisations that avoid this trap are not the ones that improve less, but the ones that define, explicitly, the point at which improvement must stop and performance must begin.

Ready? Let's talk.