Your software is a mirror of your organisation. Conway told you this in 1967. You weren't listening.

I watched this happen at a challenger bank. The CTO was paralysed by the weight of accountability to financial regulators — so scared of making a wrong call that he made no calls at all. The Head of Engineering behaved like a kid in a candy shop: migrated from GitHub to GitLab for no measurable benefit, brought in consultants to set up MongoDB and Kafka at scale before the product had users to scale for. Go-live date after go-live date was missed. Neither was the right person to be in a leadership position, and everyone in the building knew it except the two people who mattered. The CEO — who lacked the technical depth to challenge either of them on architecture decisions or delivery timelines — eventually lost his job. The people who caused the failure walked away with their titles intact. This is what an accountability vacuum looks like in practice: the person furthest from the decisions pays the price, and the people who made the decisions face no consequences because the structure made it impossible to trace cause to effect.

Look at the companies that do great work — the ones that ship products people actually want, year after year, with discipline and focus. Their team structures are as deliberate as their architectures. Small. Accountable. Clear on who owns what. Now look at your organisation. Look honestly. The bloated teams, the consensus tours, the process layered on top of process. The hiring practices lifted from a Joel Spolsky blog post written two decades ago as though the industry hasn't moved. The management positions filled by people who wanted the title rather than the responsibility. The comfortable fiction that adding a framework or a certification will solve what is fundamentally a people problem.

If you look at what's working in the industry right now, there's a good chance it's not a reflection of your org. It's a long, hard look into a reality where you're becoming irrelevant — not because you lack talent, but because your structure destroys the accountability that talent requires to function.

This is a piece about that destruction. Not the corporate poster version of accountability with stock photos and motivational quotes, but the structural, measurable, financially ruinous absence of it that defines most organisations above about fifty people. The research on this spans psychology, organisational behaviour, and management science. It has been consistent for over a century. It is not ambiguous. What follows is what it says, and what it means for anyone responsible for building teams that are supposed to deliver.

In the 1880s, a French agricultural engineer named Maximilien Ringelmann asked people to pull a rope. Alone, they pulled with maximum effort. In pairs, each person pulled slightly less hard. In groups of eight, individual effort dropped to roughly half of what each person exerted alone. Ringelmann had discovered something that would be replicated hundreds of times across the next century: individual productivity declines as group size increases.

This isn't a metaphor. It has been quantified precisely.

The decline is driven by three forces that compound as headcount grows. Motivational loss — when individual contribution is invisible, effort drops because nobody can tell who is pulling and who is coasting. Coordination loss — the overhead of synchronising multiple people consumes energy that could be spent on the task itself. And diffusion of responsibility — the psychological phenomenon where the more people share an obligation, the less any single person feels compelled to act on it.

That last one deserves attention because it scales in ways that organisational designers routinely ignore. The number of communication channels in a group follows a simple formula: n(n−1)/2. The growth is not linear. It is combinatorial.

Each channel is a potential pathway for miscommunication, misalignment, conflicting instruction, or simply wasted time. At five people, ten channels are manageable — everyone can hold the full picture in their head. At fifty, 1,225 channels means nobody can. The organisation's response to this is invariably to add process: standups, status reports, alignment meetings, RACI matrices, shared dashboards. But process doesn't reduce the channels. It adds new ones. Every standup is a broadcast channel. Every status report is an asynchronous channel. Every alignment meeting creates follow-up channels. The complexity compounds.

Frederick Brooks identified this in 1975. Fifty years later, most organisations are still solving the problem by making it worse.

In 1968, psychologists John Darley and Bibb Latané ran an experiment. Participants overheard someone having what appeared to be a seizure. When the participant believed they were the only one who could hear it, 85% intervened. When they believed four other people could also hear it, intervention dropped to 31%. The phenomenon — diffusion of responsibility — has been replicated in hundreds of studies across decades and cultures. It is one of the most robust findings in social psychology.

The mechanism is straightforward: when responsibility is shared, it is diminished. Not divided — diminished. Five people sharing responsibility do not each feel 20% responsible. They each feel something closer to 10%. Total felt responsibility in the system drops as participants increase. This is not laziness. It is a documented cognitive response to the presence of others.

Apply this to any organisation above a trivial size and the implications are stark. When a product decision requires sign-off from three stakeholders, none of them feels fully responsible for the outcome. When a project is owned by a "cross-functional team" of twelve, ownership belongs to nobody. When a failure triggers a post-mortem attended by twenty people, accountability dissolves into the collective like salt in water. Everyone was involved. Nobody was responsible.

This is not a cultural problem. It is a structural one. The architecture of the organisation creates the conditions for diffusion. And no amount of values workshops, leadership training, or motivational posters will override a cognitive response that has been observed in every human group ever studied.

J. Richard Hackman spent decades at Harvard studying team effectiveness. His research, published in Leading Teams (2002) and across dozens of peer-reviewed papers, arrived at conclusions that should have reshaped how every organisation in the world thinks about team design. They didn't, because the conclusions are uncomfortable.

Hackman found that the optimal team size for complex knowledge work is between four and six people. He was known for never permitting teams larger than six in his own classes at Harvard. His data showed a measurable decline in performance beyond that threshold — not because larger teams lacked talent, but because coordination costs and accountability diffusion overwhelmed the additional capacity.

His central insight was counterintuitive and remains widely ignored: leaders have the greatest leverage before work begins — by designing the team and its context — not by monitoring and intervening continuously. Put the right people in a small enough group with a clear enough mission, and management largely becomes unnecessary. Put the wrong people in a large group with a vague mission, and no amount of management will save it. That's Conway's law restated as an organisational design principle: the structure determines the output. Fix the structure or accept the output. There is no third option.

McKinsey's research reaches the same destination from a different starting point. Knowledge workers lose up to 28% of their working week to administrative coordination — meetings about meetings, status updates for status updates, alignment sessions that generate follow-up alignment sessions. For a team of 40 engineers at a fully-loaded cost of £90,000 per year, that is approximately £1 million annually spent on people talking about work rather than doing work. Two-thirds of leaders surveyed describe their own organisations as overly complex and inefficient — an extraordinary admission from the people who built the complexity. And the financial correlation is direct: McKinsey found that companies reporting low levels of individual complexity had the highest returns on capital employed and returns on invested capital.

The instinctive response to accountability failure is to add process. If people aren't delivering, add standups. If decisions aren't being made, add a RACI matrix. If quality is slipping, add a review gate. If communication is poor, add a weekly status report. Each intervention feels rational in isolation. Collectively, they create a system optimised for the appearance of control rather than the reality of delivery.

Ethan Bernstein's research at Harvard Business School examined this directly. He studied organisations that had implemented transparency and process initiatives designed to promote accountability, and discovered that they frequently produced the opposite effect. When employees felt constantly observed and measured, they concealed deviation from the norm rather than innovating. Experimental behaviour stopped. People optimised for looking productive rather than being productive.

Think about what this actually means. The standup that was introduced to create visibility into progress becomes a performance — a daily ritual where people report activity because activity is what gets measured, not outcomes. The dashboard that was built to surface blockers becomes a tool for demonstrating busyness. The OKR framework that was designed to align effort with strategy becomes a quarterly exercise in writing objectives that are achievable enough to never risk failure. Every transparency mechanism, designed with good intentions, creates a new incentive to perform accountability rather than practice it.

This is what happened at the challenger bank. There was no shortage of process. There were standups, there were sprint reviews, there were status reports going to the board. The CTO and Head of Engineering were visible in every ceremony. They looked accountable. The rituals of accountability were meticulously observed. But the GitLab migration nobody asked for still happened. The premature Kafka infrastructure still got built. The go-live dates still slipped. The process created the appearance of oversight while providing none of the substance — because the fundamental problem wasn't visibility, it was that the wrong people were in the seats, and no amount of process transparency will fix a leadership failure that nobody is willing to name.

The finding should alarm anyone who has recently invested in an "accountability framework": adding process to create accountability can destroy the conditions under which accountability naturally occurs. Worse, it provides cover for exactly the dysfunction it was designed to prevent. The organisation can point to the standup, the dashboard, the quarterly review and say "we have accountability." They don't. They have theatre.

This is the trap. The organisation detects an accountability problem. It responds with a structural intervention — a new process, a new role, a new tool. The intervention adds complexity. The complexity increases coordination costs. The increased coordination costs further diffuse individual ownership. Accountability declines further. The organisation detects the decline and responds with another structural intervention. The cycle repeats until the organisation is so encrusted with process that the original work — building something, shipping something, serving a customer — becomes a minority activity performed in the gaps between governance.

If the research describes the disease, the companies that don't have it confirm the diagnosis. And the pattern is always the same: small teams, clear ownership, no process theatre.

Valve Corporation is a company most business leaders have never studied because they dismiss gaming as unserious. This is a mistake. Valve operates with approximately 336 employees, generates roughly $5 billion in annual revenue, and has quietly built one of the most strategically coherent product ecosystems in the technology industry. They have no product managers. No scrum masters. No sprint ceremonies. No Jira boards. And their output over the past decade makes most organisations with fifty times their headcount look idle.

What makes Valve worth studying isn't any single product — it's the sequencing. Every major move they've made has laid the foundation for the next, in a chain of deliberate, patient, accountable bets. Steam launched in 2003 as a clunky distribution tool that was widely hated. Valve ate the criticism, iterated for years, and turned it into the dominant storefront in PC gaming — roughly 75% market share. That gave them the customer relationship and the data. Then they invested heavily in Proton, an open-source compatibility layer that allows Windows games to run on Linux. On the surface, this made no business sense — Linux desktop share was tiny. But Valve was solving a dependency problem. Their entire business ran on top of an operating system controlled by Microsoft, a competitor who could launch their own store at any time. Proton was an insurance policy.

Then the Steam Deck. Because Proton was mature enough to run thousands of Windows games on Linux, and because SteamOS was stable enough to be a real operating system, Valve could ship a handheld gaming PC that didn't need Windows — and price it aggressively because the margin was in Steam purchases over the device's lifetime, not in hardware markup. The Deck wasn't a gamble. It was a proof of investments already made. And it succeeded — building genuine consumer trust and proving that the Linux gaming ecosystem was viable. Now, in 2026, they're launching three new hardware products: a compact living-room console, a VR headset, and a controller. Each one runs on the ecosystem the previous products built. The strategy is a textbook example of sequenced, accountable product development — each step only makes sense in light of what came before and what comes after.

Contrast this with Microsoft's Xbox division, which is currently experiencing the accountability vacuum in real time. Hardware revenue dropped 33% year-on-year in the most recent quarter. They've gone through a CEO change. Their AI gaming assistant, launched as a centrepiece feature barely a year ago, has already been killed. Game Pass price increases drove cancellations, which demands more content to justify the higher price, but they're simultaneously laying off studios and cancelling projects. Their developers report that Xbox certification is slow and aggressive, and even major publishers struggle with basic platform tooling. This is what happens when an organisation with 228,000 employees tries to compete with one that has 336. The accountability is diluted across so many layers that nobody owns the outcome, so nobody fixes the structural problems, so the structural problems compound.

Apple under Jobs operated on the same principle as Valve but through different mechanisms. Jobs didn't hide from hierarchy — he was explicit about it. Small teams, clear leaders, direct accountability. His observation that the best managers are the ones who don't want the job but know they have to do it is the most concise description of accountable leadership ever articulated. He wasn't describing a framework. He was describing a type of person — and insisting that finding that person matters more than any process you wrap around the wrong one. Apple's product sequencing was equally deliberate: iPod proved digital distribution, iTunes built the customer relationship, iPhone leveraged both. Each product was accountable to the ecosystem, not to a quarterly roadmap.

What makes the Apple example more than historical nostalgia is that the structure survived Jobs. Tim Cook maintained the functional organisation model for fifteen years — departments organised by expertise, each led by a single SVP who owns their domain and reports directly to the CEO. No product divisions. No committees. The same architecture Jobs described in that red chair in 2010, running a company that grew from $65 billion to $4 trillion in market cap. And in April 2026, Apple announced that John Ternus — a hardware engineer who joined in 2001, spent 25 years building products, and rose through the ranks because his competence demanded it — will become CEO in September. Not a finance leader. Not an operations specialist. An engineer who knows how to ship. That's three CEO transitions and the structural principle hasn't changed, because it works. The idea has outlived the man who planted it.

Look at those numbers. Valve generates nearly fifteen million dollars per employee — more than six times Apple and nearly fourteen times Microsoft. Not because they've found some magic framework. Because every person is visible, every person is accountable, and the organisation refuses to add complexity to compensate for bad decisions. Being private means Newell never has to perform for analysts or manufacture narratives for a stock price. He can make a decade-long bet on Proton with no direct revenue attached because nobody is asking him to justify it in a quarterly earnings call. That patience — the freedom to sequence products over years rather than quarters — is a structural advantage that public markets systematically undervalue and that bloated organisations structurally cannot replicate.

The lesson from both companies is not about flatness or hierarchy. It's about what they refuse to do. Neither solves people problems with process. Neither compensates for the wrong person in the wrong seat by adding a framework around them. Neither diffuses decision-making across committees to reduce risk. Both accept that a wrong hire is a wrong hire and that the solution is to fix the hire, not to build scaffolding around the gap. And both would rather ship nothing than ship something nobody owns.

The financial argument is not subtle. If a team of 40 is delivering what a team of 8 could deliver — and the research strongly suggests that many organisations are operating at exactly this ratio — then 32 salaries are being spent on coordination overhead, social loafing, and the organisational theatre required to maintain the illusion of productive collaboration. At UK market rates for experienced engineers, that is somewhere between £2.5 million and £4 million per year in waste. Per team.

Multiply across an enterprise with dozens of such teams and the figure becomes existential. Not in the sense that the company will fail tomorrow, but in the sense that it is carrying costs that a leaner, more accountable competitor does not carry. When that competitor arrives — and in the age of AI-augmented small teams, it will arrive sooner than most incumbents expect — the bloated organisation will discover that its process infrastructure is not an asset but a liability. You cannot pivot quickly when every decision requires a consensus tour across six departments and a change advisory board.

The talent cost is equally severe, though harder to measure. Good people — the ones you most need to retain — are the first to recognise an accountability vacuum. They see that their effort is invisible in a group of twenty. They see that decisions they could make in an hour require a week of stakeholder alignment. They see that the person who should have been managed out six months ago is still there because nobody owns the decision to act. And they leave. Not loudly, not dramatically, but steadily, until the organisation is left with the people who are comfortable in the vacuum — the ones who have optimised for survival in a system that rewards presence over output.

If you want a case study in accountability failure at industrial scale, look no further than the last decade of "digital transformation." The phrase itself became a talisman — a magic word that justified enormous budgets, executive hires, and multi-year programmes. The results have been devastating.

Seventy percent. In any other field, a 70% failure rate would trigger an existential reckoning. In digital transformation, it barely registers — because the people responsible for the failures are rarely the people who bear the consequences. The consultancy that designed the programme moves on. The CTO who championed it has already changed roles. The transformation lead's LinkedIn profile describes a "major digital initiative" without mentioning that it delivered nothing. Accountability, once again, dissolves into the structure.

But the failure pattern is what matters here, because it maps directly onto the accountability vacuum. McKinsey's own research states it plainly: successful digital transformation is 20% technology and 80% organisational change. Yet most companies invert that ratio entirely, spending the majority of budget on technology while treating the people and structural dimension as an afterthought. They thought the transformation didn't extend beyond the server. Buy the platform, migrate the data, implement the tool. Transformation complete. Except nothing actually changed — because the organisation that was dysfunctional before the technology arrived is still dysfunctional after it. The same bloated teams, the same diffused ownership, the same consensus-paralysis, now running on a more expensive platform.

That statistic should end every argument about whether organisational size and structure affect outcomes. Small companies — where individuals are visible, where ownership is clear, where there's nowhere to hide — succeed at transformation nearly three times the rate of large ones. Not because they have better technology. Because they have fewer layers between a decision and its consequence. When the person who chose the platform is also the person who has to make it work, the choice tends to be better.

The worst-performing sectors are instructive too. McKinsey found that traditional industries — oil and gas, automotive, infrastructure, pharmaceuticals — achieved digital transformation success rates as low as 4% to 11%. These are industries with deep management hierarchies, extensive governance structures, and decades of accumulated process. They are also, not coincidentally, industries where the accountability vacuum is most entrenched. The people leading the transformation were too conservative, too insulated from consequences, and too anchored to a mental model where "transformation" meant upgrading systems rather than restructuring how humans make decisions and ship work.

The global cost of this failure is estimated at $2.3 trillion. That is not a rounding error. It is the price of an entire industry's inability to confront the fact that technology does not fix broken organisations — it amplifies them. Conway's law, again: if your organisation is structured for paralysis, your digital transformation will produce a more expensive, more technologically sophisticated form of paralysis.

The research points in one direction, and it is not toward a new framework. It is toward a set of decisions that most leaders find genuinely difficult to make because they require the very thing the accountability vacuum has trained out of them: personal responsibility for an uncomfortable call.

It requires making teams small enough that individuals are visible. Hackman says four to six. Brooks' formula explains why. Ringelmann's data quantifies the cost of ignoring it. This is not a suggestion. If your team is larger than six and you're wondering why nothing ships, you have your answer. Split it. Give each smaller group clear, distinct ownership. Accept that this means fewer people in fewer meetings feeling important.

It requires putting one name on every outcome. Not a team name. Not a squad name. A person. Diffusion of responsibility is not a behaviour you can coach away — it is a cognitive inevitability in groups. The only structural defence is to make ownership explicit and singular. One person owns the decision. One person answers for the result. Everyone else is input, not authority.

It requires stripping out every process that exists to compensate for having the wrong people. Ask yourself, honestly, about every standup, every status report, every review gate: would this be necessary if the team were four excellent people who trusted each other? If the answer is no, you don't have a process gap. You have a people gap. And you're paying for a process to avoid confronting it.

It requires leaders who don't want the job but know they have to do it. The EOS framework calls it "right person, right seat." Jobs described it as finding people who understand they have to lead because nobody else will and the work demands it. These are people who shield teams from organisational noise, who make the call when consensus fails, who act when someone isn't performing. They are, in the language we use at uRadical, vanishingly rare — and no certification programme in the world will produce them. You find them, or you don't. There is no process that substitutes for their absence.

And it requires measuring output, not activity. If your metrics track velocity, story points, tickets closed, or meetings attended, you are measuring the organisation's ability to perform work-shaped activities, not its ability to deliver outcomes. You have built a system that rewards the appearance of productivity. And you will get exactly what you've incentivised: a building full of busy people producing nothing of consequence.

None of this research is new. Ringelmann's work is from the 1880s. Conway published in 1968. Brooks' law is from 1975. Darley and Latané published in 1968. Hackman's findings have been available for over twenty years. McKinsey has been publishing on organisational complexity for more than a decade. The data is not the problem.

The problem is that acting on it requires the very thing the data describes as absent: someone willing to take personal responsibility for an uncomfortable structural decision. Someone willing to say this team is too big, this person is in the wrong seat, this process exists because we're afraid to have a difficult conversation, and this hiring practice is two decades out of date because we never questioned it.

Finding the right people isn't as simple as finding somebody who can quote the latest management speak or believes in outdated hiring rituals because they read a blog post from a thought leader in 2004 and never updated their thinking. It requires looking honestly at what is working in the industry — not what's comfortable, not what's familiar — and accepting that the answer might be a hard look at a reality where your organisation has become irrelevant. Not because the market moved. Because your structure made you unable to move with it.

If you read this and recognised your own organisation — in the consensus tours, in the diffused ownership, in the process layering, in the teams that are too large staffed with people you're not sure should be there — then you already know what needs to change. The question is whether anyone in your building is accountable for making it happen.

If the answer is nobody, you have the problem this article describes. And no framework will fix it for you.