Every department hitting its numbers is not evidence against a silo problem. In most organisations, it is closer to what a silo problem actually looks like from the inside.
This is the specific insight behind Eliyahu Goldratt’s Theory of Constraints, developed in the manufacturing context of his 1984 book The Goal and since generalised well beyond it. Goldratt’s central claim directly contradicted standard management practice at the time: that improving the whole system is not the sum of improving each of its parts. A change to most variables in an organisation, he argued, has only a small effect on overall performance, because performance is typically governed by a small number of genuine constraints rather than by the aggregate health of every department. The corollary is less comfortable than the claim itself. A business can optimize every department individually — hit every departmental target, satisfy every departmental KPI — and see almost no improvement at the level that actually matters, because the parts were never the constraint. The connections between them were.
Gillian Tett’s 2015 book The Silo Effect documents this pattern across eight organisations, and the contrast between its failure cases and its recovery cases is instructive. At Sony, Tett describes divisions operating like separate “octopus pots” — each unit sealed off, optimising its own performance, occasionally competing against sister divisions rather than coordinating with them, while the company as a whole lost ground it might otherwise have defended. UBS’s risk functions, in Tett’s account, suffered from a related problem: departments each managing their own slice of risk competently, with no single view of how those slices combined. Cleveland Clinic sits on the other side of the same book. Facing a structure organised around medical specialties — cardiology, surgery, nephrology, each excellent on its own terms — the hospital reorganised entirely around patient conditions instead, building “institutes” that grouped whichever specialists a given condition actually required into one coordinated unit. The individual specialties did not become less skilled. What changed was that the system connecting them was deliberately redesigned, rather than left to whichever cross-functional relationships happened to form informally.
[ Read More: Process and System Design Services ]
Harvard Business Review’s research into what its authors termed “collaborative overload” offers a useful picture of what happens while that redesign work goes undone. Rob Cross, Reb Rebele, and Adam Grant found that time spent by managers and employees in collaborative activities — meetings, calls, email — had increased by 50% or more over roughly two decades, and that at many companies, the proportion of a working day consumed by these activities had reached around 80%. Their more striking finding concerned where that collaborative burden actually falls: 20% to 35% of an organisation’s value-added collaboration comes from just 3% to 5% of its employees. In practice, that describes a small number of people functioning as informal connective tissue between departments that have no formal mechanism for reconciling with one another — and the same research found that these individuals report disproportionately low engagement and satisfaction, precisely because the coordination they are performing was never designed as anyone’s actual role.
With that framing established, a small number of signs are worth checking for specifically.
Departmental KPIs are consistently met, but a company-level outcome — revenue, margin, retention — stays flat or declines regardless. This is Goldratt’s point made concrete: the parts are being optimised individually while the connecting system remains unmanaged.
The same two or three names appear on nearly every cross-functional email thread, not because their role formally requires it, but because they are the only people who can translate context from one department into terms another will understand. This is the “3 to 5%” finding from the HBR research, visible inside a single organisation.
A basic company-wide question — why did revenue move last quarter, what actually happened with a specific customer segment — cannot be answered from an existing document. It requires a meeting to assemble the answer from separately maintained reports that were never built to speak to one another.
Two departments use the same word to mean different things, and have never formally reconciled it. A “qualified lead” in marketing and a “qualified lead” in sales are common offenders, but the pattern recurs anywhere two functions hand work to each other without a shared definition of what is being handed over.
[ Read More: Why Most Business Consultants Solve the Wrong Problem ]
New initiatives consistently stall at the handoff points between teams rather than within any single team’s execution — Sony’s octopus pots, in miniature, repeated at every project boundary.
Coordination activity — meetings, status syncs, cross-team updates — has grown noticeably faster than actual output over the same period. Growth in the former is not a sign of a more collaborative culture. It is frequently a sign of a system quietly compensating, through human effort, for coordination that was never designed into the business’s actual structure.
None of this is especially actionable as diagnosis alone, so it is worth setting out what actually follows from recognising it.
The first step is making the silo visible with numbers rather than accumulated frustration. This is, not incidentally, why silo problems tend to surface for the first time during a Detailed Project Report rather than during ordinary operations. A properly conducted DPR is one of the few exercises that requires every function — financial, operational, market-facing, systems — to be reconciled against a single, coherent narrative of the business, rather than reported separately and left uncompared. Running that kind of cross-functional reconciliation, even as a standalone exercise, is usually what converts “it feels like departments aren’t talking” into a specific, defensible list of where they actually are not.
The second step is resisting the instinct to treat the whole organisation as equally siloed. Goldratt’s framework is explicit on this point: only a small number of constraints govern overall performance, which means most connection points between departments are not where the damage is occurring. The DPR-style reconciliation exercise should produce a short list of specific handoffs — the place where marketing’s numbers stop matching sales’, the point where product and operations disagree about capacity — rather than a general mandate to “improve collaboration” everywhere at once.
The third step is redesigning the system at those specific points, which is a materially different exercise from adding another meeting. Cleveland Clinic’s fix was not a new coordination committee layered on top of the existing specialty structure; it was restructuring which people were formally grouped together in the first place, so the coordination happened by default rather than by goodwill. This is what business systems design work is for — not managing the symptoms of a broken handoff through more status updates, but rebuilding the handoff itself so it no longer depends on three overloaded people remembering to loop each other in.
[ Read More: Brand Management Services: Build a Strong, Clear, and Consistent Brand ]
The fourth step is assigning a named owner to the connection point itself, distinct from the departmental managers on either side of it. Most organisations assign clear ownership within departments and almost none assign ownership of what happens between them, which is precisely the gap the earlier signs describe. A cross-functional process without an accountable owner tends to default back to whichever individual happens to have the most context — reproducing the 3-to-5% concentration problem the HBR research documented, rather than resolving it.
The fifth step is measuring success at the level Goldratt actually cares about. If the redesign is working, the signal will not be that every department’s dashboard turns slightly greener. It will be that the company-level number — the one that stayed flat while every department looked fine — finally starts moving, because the constraint that mattered was never inside a department to begin with.
